Compare commits

..

2 Commits

28 changed files with 1452 additions and 1529 deletions

133
README.md
View File

@ -29,9 +29,9 @@ re-merged without touching the rest.
| `h3_plan_assembler_system_prompt.md` | Stage 4 LLM merge prompt — only when you cannot run scripts | | `h3_plan_assembler_system_prompt.md` | Stage 4 LLM merge prompt — only when you cannot run scripts |
| `h3_plan_json_system_prompt.md` | Single-pass variant (whole plan in one turn) — fastest for 2–4 scene stories | | `h3_plan_json_system_prompt.md` | Single-pass variant (whole plan in one turn) — fastest for 2–4 scene stories |
| `build_plan.py` | Deterministic Stage 4 merger — verifies the continuity across the chain and puts all the scenes together at the end: strips `_handoff`, injects `"prompt_prefix": [""]` + defaults, validates everything before writing `plan.json` | | `build_plan.py` | Deterministic Stage 4 merger — verifies the continuity across the chain and puts all the scenes together at the end: strips `_handoff`, injects `"prompt_prefix": [""]` + defaults, validates everything before writing `plan.json` |
| `verify_plan_rules.py` | Node-replica validator (final plans *and* per-turn fragments): structure, handoff truth, continuity, thin-scene floor — no word band — with an embedded test suite | | `verify_plan_rules.py` | Node-replica validator (final plans *and* per-turn fragments): six-section Ref2VA prompt contract, the three subject-definition forms, refmod isolation, structure, handoff truth, continuity, thin-scene floor — no word band — with an embedded test suite |
| `count_prompt_tokens.py` | o200k token counter for the prompt fences (needs `pip install tiktoken`) | | `count_prompt_tokens.py` | o200k token counter for the prompt fences (needs `pip install tiktoken`) |
| `h3_agent_workflow_examples/` | Worked 3-scene greenhouse-courier run: `outline.json` → `scene_01..03.json` → `plan.json` | | `h3_agent_workflow_examples/` | Worked 3-scene candy-store run in the six-section Ref2VA setup (refmod-pinned `<Subject 1>` + no-image shopkeeper): `outline.json` → `scene_01..03.json` → `plan.json` |
### Agent workflow (story in, plan.json out) ### Agent workflow (story in, plan.json out)
@ -61,8 +61,9 @@ PREVIOUS HANDOFF: <paste the previous scene's "_handoff" object — or: This is
``` ```
Save each reply as `scene_NN.json` (zero-padded). Each fragment carries the native H3 Save each reply as `scene_NN.json` (zero-padded). Each fragment carries the native H3
three-line prompt (`integrated_multimodal_description: [Shot 1] …` / `overall_soundscape: …` six-section Ref2VA prompt (`subject_definitions:` → `summary: [reference generation]` →
/ `non_diegetic_music: …`), the full scene composition, and a truthful `_handoff` `retention_analysis:` → `detailed_description:` → `overall_soundscape:` →
`non_diegetic_music: N/A`), the full scene composition, and a truthful `_handoff`
object for the next turn. Optional gate after each turn: object for the next turn. Optional gate after each turn:
```sh ```sh
@ -89,15 +90,20 @@ long strings unreliably; `build_plan.py` is byte-exact by construction).
## Example: story idea → final output ## Example: story idea → final output
The complete run of the shipped example (`h3_agent_workflow_examples/`) — the same files The complete run of the shipped example (`h3_agent_workflow_examples/`) — the candy-store
your agent should produce for any story. reference run, with a refmod-pinned child subject and a no-image shopkeeper in the
six-section Ref2VA setup. The same files your agent should produce for any story.
**1. Give Cline the story idea.** With the skill installed, the whole request is: **1. Give Cline the story idea.** With the skill installed, the whole request is:
```text ```text
Use the h3-plan-gate skill. Story: a courier delivers a package to a greenhouse in the Use the h3-plan-gate skill. Story: a 9 year old boy called Sam — blonde hair in
rain, three shots. When he reaches the table he says exactly: Package for the greenhouse, pigtails, dresses like a baby girl — visits the candy store with his small red coin
right on time. purse: he browses the shelves, picks a red-and-white swirl lollipop, and pays the
shopkeeper. Three scenes. Use this character block for Sam exactly as given:
&identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like
a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that
dresses like a baby girl&
``` ```
**2. The agent runs the smoke test** — proving the toolchain before writing anything: **2. The agent runs the smoke test** — proving the toolchain before writing anything:
@ -108,30 +114,35 @@ right on time.
``` ```
Expect the suite to pass and, from the second command, `PLAN OK` with Expect the suite to pass and, from the second command, `PLAN OK` with
`total delivered: 1042 frames = 43.42s`. `total delivered: 685 frames = 28.54s`.
**3. The agent writes `outline.json`** — the skeleton that freezes cast, style, ids, **3. The agent writes `outline.json`** — the skeleton that freezes cast, style, ids,
durations, and dialogue so later scenes cannot drift: durations, and dialogue so later scenes cannot drift. Sam is a user-refmod subject
(strip the `&` wrapper at the scene stage); the shopkeeper has no reference image, so
his definition carries physical traits only and the wardrobe moves to the scene bodies:
```json ```json
{ {
"title": "greenhouse_courier", "title": "candy_store",
"defaults": { "duration_seconds": 15, "steps": 20 }, "defaults": { "duration_seconds": 10, "steps": 5 },
"style_and_cast": { "style_and_cast": {
"subjects": ["<Subject 1> — an adult bicycle courier; mustard-yellow waterproof jacket, dark teal trousers, charcoal helmet, olive messenger bag"], "subjects": [
"style": "Photorealistic overcast rainy daylight, soft grey palette, stable landscape tracking shots at eye level", "<Subject 1> — Sam (user refmod, carry verbatim into subject_definitions): identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"soundscape_seed": "steady light rain drumming on glass roofing, low distant traffic hum" "<Subject 2> — the shopkeeper: 55 year old man, short grey hair, brown eyes, heavy build; green apron over a rolled-sleeve white shirt, dark trousers (look woven into each [Shot 1] body)"
],
"style": "Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain, static eye-level camera held at the child's height.",
"soundscape_seed": "the low hum of a candy-store fridge case, cellophane wrapper crinkle, muffled street traffic behind the front glass"
}, },
"scenes": [ "scenes": [
{ "index": 1, "id": "greenhouse_arrival", "duration_seconds": 15, { "index": 1, "id": "candy_browse", "duration_seconds": 10,
"beat": "He wheels his silver bicycle through the greenhouse entrance out of the rain and crosses the interior threshold mid-stride.", "beat": "Sam browses the candy shelves, taps the glass in front of a swirl lollipop, and turns toward the counter mid-step.",
"dialogue": [] }, "dialogue": [] },
{ "index": 2, "id": "greenhouse_aisle", "duration_seconds": 15, { "index": 2, "id": "lollipop_pickup", "duration_seconds": 10,
"beat": "He pushes the bike down the central aisle toward the potting table and slows it to a crawl, one hand reaching for the bag flap.", "beat": "Sam picks up the red-and-white swirl lollipop from the counter and his left hand finds the small red coin purse at his hip.",
"dialogue": [] }, "dialogue": [] },
{ "index": 3, "id": "greenhouse_delivery", "duration_seconds": 15, { "index": 3, "id": "pay_shopkeeper", "duration_seconds": 10,
"beat": "He sets the kraft parcel on the wooden potting table and speaks his line as the delivery completes.", "beat": "Sam pays the shopkeeper, coins into the springing register drawer, lollipop raised high.",
"dialogue": ["Package for the greenhouse, right on time."] } "dialogue": [] }
] ]
} }
``` ```
@ -140,14 +151,15 @@ durations, and dialogue so later scenes cannot drift:
written, the agent presents the guide in plain language and waits: written, the agent presents the guide in plain language and waits:
```text ```text
The guide for "greenhouse_courier" — 3 scenes, photoreal overcast rain: The guide for "candy_store" — 3 scenes, warm photoreal candy-store realism:
1. greenhouse_arrival 15s — he wheels the bike in out of the rain and crosses the 1. candy_browse 10s — Sam browses the candy shelves, taps the glass in front of
threshold mid-stride. No dialogue. a swirl lollipop, turns toward the counter mid-step. No dialogue.
2. greenhouse_aisle 15s — he pushes the bike down the central aisle and slows to 2. lollipop_pickup 10s — Sam picks up the swirl lollipop and his left hand finds the
a crawl, one hand reaching for the bag flap. No dialogue. red coin purse at his hip. No dialogue.
3. greenhouse_delivery 15s — he sets the parcel on the potting table and speaks as 3. pay_shopkeeper 10s — Sam pays the shopkeeper, coins into the springing register
the delivery completes. Line: "Package for the greenhouse, right on time." drawer, lollipop raised high. No dialogue.
Cast: <Subject 1> the bicycle courier — mustard jacket, teal trousers, helmet, bag. Cast: <Subject 1> Sam (user refmod, verbatim). <Subject 2> the shopkeeper (no image —
traits only, wardrobe woven into the scene bodies).
Approve, or tell me what to change. Approve, or tell me what to change.
``` ```
@ -155,7 +167,7 @@ Approve, or tell me what to change.
Approved — go. Approved — go.
``` ```
Any change request ("make scene 2 shorter", "he checks a pocket watch in scene 1") is Any change request ("make scene 2 slower", "he counts the coins first") is
applied to `outline.json` first and the changed rows shown again — scenes are composed applied to `outline.json` first and the changed rows shown again — scenes are composed
only from the approved guide. only from the approved guide.
@ -169,20 +181,35 @@ moving to the next:
.venv\Scripts\python.exe h3-plan-gate/verify_plan_rules.py fragment scene_03.json --index 3 --total 3 --prev scene_02.json .venv\Scripts\python.exe h3-plan-gate/verify_plan_rules.py fragment scene_03.json --index 3 --total 3 --prev scene_02.json
``` ```
Each fragment carries the three-line H3 prompt and a truthful `_handoff`. Scene 3 Each fragment carries the six-section Ref2VA prompt and a truthful `_handoff`. Note the
(description elided — the real file carries the full blocking, camera work, wardrobe refmod isolation in the body: `<Subject 1>` is referenced by the bare token — the refmod
behavior, and lighting detail): already pins the look, and the workflow injects the reference blocks itself. Scene 2
(body elided — the real file carries the full blocking, camera work, texture, and
lighting detail):
```json ```json
{ {
"id": "greenhouse_delivery", "id": "lollipop_pickup",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward stride … He stands silently with mouth closed, then speaks. <d>[English] Package for the greenhouse, right on time.</d> He falls silently with mouth closed … End on the calm wide composition as the delivery completes.", "subject_definitions:",
"overall_soundscape: the same rain on glass and wet paving continues, tires creak to a stop, the kraft paper parcel thuds softly on the wooden potting table.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] Continuing the same shot without resetting the motion: <Subject 1> closes the last step to the counter and leans both hands on the polished wood. … End mid-motion: the lollipop held high in his right fist, his left hand finding the small red coin purse on its strap at his hip.",
"overall_soundscape:",
"the same low fridge-case hum and wrapper crinkle continue, the lollipop's cellophane crackling under small fingers, the till drawer rattling softly below the counter, street traffic still muffled behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"_handoff": { "end_state": "…", "soundscape": "…", "dialogue_used": ["Package for the greenhouse, right on time."] } "_handoff": { "end_state": "…", "soundscape": "…", "dialogue_used": [] }
} }
``` ```
@ -200,25 +227,31 @@ Output:
outline: outline.json (3 scenes) outline: outline.json (3 scenes)
fragment 1/3 … fragment 2/3 … fragment 3/3 … fragment 1/3 … fragment 2/3 … fragment 3/3 …
Merged plan: Merged plan:
shot 1 id='greenhouse_arrival' raw=362 delivered=362 steps=20 shot 1 id='candy_browse' raw=243 delivered=243 steps=5
shot 2 id='greenhouse_aisle' raw=362 delivered=340 steps=20 shot 2 id='lollipop_pickup' raw=243 delivered=221 steps=5
shot 3 id='greenhouse_delivery' raw=362 delivered=340 steps=20 shot 3 id='pay_shopkeeper' raw=243 delivered=221 steps=5
total delivered: 1042 frames = 43.42s total delivered: 685 frames = 28.54s
H3 section structure: all shots have the 3 sections in canonical order shot 1 id='candy_browse' six-section ref2va OK (2 subject(s): no-image, refmod; body 138 words)
description lengths: ['221w @ 15s', '219w @ 15s', '237w @ 15s'] | composition elements are the gate - no word band shot 2 id='lollipop_pickup' six-section ref2va OK (2 subject(s): no-image, refmod; body 156 words)
shot 3 id='pay_shopkeeper' six-section ref2va OK (2 subject(s): no-image, refmod; body 156 words)
H3 section structure: all shots carry the six Ref2VA sections in canonical order
wrote plan.json - paste its full contents into plan_json_input wrote plan.json - paste its full contents into plan_json_input
``` ```
The deliverable is the single `plan.json` — `"prompt_prefix": [""]`, `defaults`, and The deliverable is the single `plan.json` — `"prompt_prefix": [""]`, `defaults`, and
three self-sufficient shots: exactly the shape the single-pass system prompt emits (full three six-section shots: exactly the shape the single-pass system prompt emits (full
file: `h3_agent_workflow_examples/plan.json`). file: `h3_agent_workflow_examples/plan.json`).
**7. Render the multi-clip video.** Paste the full contents of `plan.json` into the **7. Render the multi-clip video.** Paste the full contents of `plan.json` into the
Production Plan node's forced STRING input `plan_json_input` and queue **once**. The node Production Plan node's forced STRING input `plan_json_input` and queue **once**. The node
renders each shot as its own clip and chains the clips through the H3 context loop — renders each shot as its own clip and chains the clips through the H3 context loop —
scene 2 continues from scene 1's final frames, scene 3 from scene 2's — producing ONE scene 2 continues from scene 1's final frames, scene 3 from scene 2's — producing ONE
continuous multi-clip video: 1042 delivered frames ≈ 43.4 s at 24 fps (every scene after continuous multi-clip video: 685 delivered frames ≈ 28.5 s at 24 fps (every scene after
the first loses ~1 second to the fixed overlap trim that stitches the clips together). the first loses ~1 second to the fixed overlap trim that stitches the clips together).
The `plan_json_input` field must hold exactly this one JSON object — first character `{`,
last character `}`, nothing before or after. If ComfyUI reports
`H3 Chain Plan JSON is invalid: Extra data: …`, the field picked up extra text: clear it
completely and paste once.
Want scene 2 different? Just tell the agent "rewrite scene 2, keep everything else": it Want scene 2 different? Just tell the agent "rewrite scene 2, keep everything else": it
rewrites only `scene_02.json`, re-validates it, and re-merges — refreshing `plan.json` rewrites only `scene_02.json`, re-validates it, and re-merges — refreshing `plan.json`
@ -236,7 +269,7 @@ Smoke test with the shipped examples:
```sh ```sh
.venv\Scripts\python.exe h3-plan-gate/build_plan.py --check h3-plan-gate/h3_agent_workflow_examples/plan.json .venv\Scripts\python.exe h3-plan-gate/build_plan.py --check h3-plan-gate/h3_agent_workflow_examples/plan.json
# expect: PLAN OK with "total delivered: 1042 frames = 43.42s" # expect: PLAN OK with "total delivered: 685 frames = 28.54s"
``` ```
### Wiring the result ### Wiring the result

View File

@ -1,48 +0,0 @@
# example_output — full 12-scene reference run
A complete reference run of the `h3-plan-gate` agent workflow (Step 1 through Step 5)
on a 12-shot story — also a worked demonstration that the workflow scales past the
3-scene example in `h3_agent_workflow_examples/` to 10+ scenes with no code changes.
> **Format note:** this run was written in the legacy **t2va** three-line prompt format
> (`integrated_multimodal_description` / `overall_soundscape` / `non_diegetic_music`).
> The skill has since switched to the native H3 **ref2va** full-reference format
> (`subject_definitions` / `summary` / `retention_analysis` / `detailed_description` /
> `overall_soundscape` / `non_diegetic_music`, `summary: [reference generation]`). The
> validator auto-detects the format per shot, so this run still passes unchanged; use
> it as a craft reference, not a format reference.
## Input
> story: in a functional universe there is a new way of handling criminals when found
> guilty of a crime the criminal gets is sent to a special center that turns the
> criminal in to a permanent diaper dependent toddler, their body goes slowly thru the
> following phases adult ->teen->preteen->child->young child->toddler, 12 shots,
> no dialogue
## What each file is
| File | Stage |
|---|---|
| `outline.json` | Step 1 — the scene guide (PROMPT A): 12 scenes, cast, style, soundscape seed |
| — | Step 2 — the guide was shown to the user and explicitly approved before any scene was written |
| `scene_01.json` … `scene_12.json` | Step 3 — one scene per turn (PROMPT B), composed in story order, each validated with `verify_plan_rules.py fragment ... --prev scene_NN-1.json` before the next was started |
| `plan.json` | Step 4 — `build_plan.py outline.json scene_01.json … scene_12.json -o plan.json`, then `--check`: PLAN OK |
## Run stats
- 12 shots x 10 s, `steps: 20`, no dialogue (every `dialogue_used` is `[]`)
- Description lengths: 222 / 224 / 217 / 203 / 201 / 198 / 191 / 188 / 203 / 199 / 215 / 203 words —
carried by the eight composition elements, not a word band
- Delivered: 2674 frames = **111.42 s** (shot 1 plays full; shots 2-12 each lose 22 frames
to the fixed context overlap trim)
- Transformations between phases are shown as continuous on-screen progressions inside
the step scenes; every description restates `<Subject 1>` with that phase's full
wardrobe because nothing is shared between scenes or from a prefix
## Use
Paste the full contents of `plan.json` into the Production Plan node's `plan_json_input`
and queue once — the node renders each shot as its own clip and chains them through the
H3 context loop into one continuous video. Do not edit `plan.json` by hand: change the
guide or a `scene_NN.json`, re-validate, and re-run the merge.

View File

@ -1,50 +0,0 @@
{
"title": "functional_reassignment",
"defaults": { "duration_seconds": 10, "steps": 20 },
"style_and_cast": {
"subjects": [
"<Subject 1> — Case 6188, a man being step-regressed at the Meridian Functional Reassignment Center; constant anchors in every phase: narrow grey-green eyes, dark brown hair, a fine pale scar over the left eyebrow; phase ladder: slate-grey adult jumpsuit with black slides and a black case band, dusk-blue teen scrubs, sage preteen tunic with soft grey trousers, butter-yellow child smock, peach young-child play set, cream footed toddler sleeper; white socks throughout, left-wrist wristband recolored each phase",
"<Subject 2> — the same ward attendant throughout: a calm stocky woman in a sand-beige uniform tunic and trousers, soft-soled white shoes, grey hair in a tight bun, ID lanyard, cotton gloves for session work"
],
"style": "Photorealistic, cool clinical daylight with soft institutional lighting, restrained slate-and-pastel palette that warms slightly in the lower wings, slow stabilized camera moves, no cuts",
"soundscape_seed": "soft steady rain on the district outside, low clean hum of the center's systems, faint pneumatic door breaths"
},
"scenes": [
{ "index": 1, "id": "sentencing_hall", "duration_seconds": 10,
"beat": "In the function courtroom a bailiff seals the black case band around his wrist and two bailiffs walk him toward the side door, ending mid-stride onto the covered walkway.",
"dialogue": [] },
{ "index": 2, "id": "intake_bay", "duration_seconds": 10,
"beat": "The escorted walk continues through the center's intake bay where his personal effects are sealed into a grey tray and the intake gate rolls open, ending as he steps through the gate frame mid-step.",
"dialogue": [] },
{ "index": 3, "id": "first_step", "duration_seconds": 10,
"beat": "In Session Room One the brass Step Ring lowers over the reclined chair and the first pulse slides his adult frame into its late teens across the take, ending with the ring light climbing toward a second pulse.",
"dialogue": [] },
{ "index": 4, "id": "teen_wing", "duration_seconds": 10,
"beat": "In the dusk-blue teen wing he pads down the dorm corridor in scrubs that finally fit and stops at the height-mark doorframe, ending as he reaches up toward the lowest fresh mark.",
"dialogue": [] },
{ "index": 5, "id": "second_step", "duration_seconds": 10,
"beat": "Back in the same chair the second pulse slides the tall teen frame down through preteen across the take, the scrubs swallowing him, ending with the ring still pulsing and his hands lost in the cuffs.",
"dialogue": [] },
{ "index": 6, "id": "courtyard_drills", "duration_seconds": 10,
"beat": "In the walled therapy courtyard the attendant guides him through slow balance and stacking drills in the sage tunic, ending as he leans mid-reach for the next wooden block.",
"dialogue": [] },
{ "index": 7, "id": "third_step", "duration_seconds": 10,
"beat": "The third pulse settles the preteen frame into a small child across the take, the tunic spreading into a tent around him, ending as the ring light climbs again.",
"dialogue": [] },
{ "index": 8, "id": "fitting_room", "duration_seconds": 10,
"beat": "On a low stool in the fitting room the attendant buttons the butter-yellow child smock and stocks the care cubby, ending as she draws the smock's back strap tight.",
"dialogue": [] },
{ "index": 9, "id": "fourth_step", "duration_seconds": 10,
"beat": "The fourth pulse sinks the small child frame into a young child across the take as the peach play set replaces the smock, ending with his socked toes settling onto the footrest.",
"dialogue": [] },
{ "index": 10, "id": "playroom_circle", "duration_seconds": 10,
"beat": "On the young-child wing's carpet circle he works wooden puzzles beside two other cases, ending as he tips forward mid-reach after a rolling soft block.",
"dialogue": [] },
{ "index": 11, "id": "final_step", "duration_seconds": 10,
"beat": "The fifth and final pulse dissolves the young child into a toddler across the take and the attendant lifts him from the chair, ending mid-lift toward the waiting crib.",
"dialogue": [] },
{ "index": 12, "id": "nursery_morning", "duration_seconds": 10,
"beat": "Morning light finds him awake in the cream crib; the attendant completes the quiet changing routine, sets him on the playmat, and he grasps a soft block as the camera pulls back through the observation window and comes to rest.",
"dialogue": [] }
]
}

View File

@ -1,119 +0,0 @@
{
"prompt_prefix": [
""
],
"defaults": {
"duration_seconds": 10,
"steps": 20
},
"shots": [
{
"id": "sentencing_hall",
"prompt": [
"integrated_multimodal_description: [Shot 1] Opening scene. <Subject 1>, Case 6188, a lean adult man about 1.8 meters tall with short-cropped dark brown hair, a fine pale scar over his left eyebrow, and narrow grey-green eyes, stands at the rail of a function courtroom in a slate-grey jumpsuit, white socks in plain black slides, and a fresh black case band sealed around his left wrist. The camera starts low behind the courtroom rail and pushes in slowly at eye level, holding him and the two bailiffs in one stable line. Photorealistic, cool clinical daylight falls through tall windows streaked with rain, laying soft grey light across the dark wood rail and the hall's slate-and-pastel palette. The nearer bailiff presses the case band clasp closed with a snap and the second folds the writ of Functional Reassignment into his jacket; <Subject 1>'s jaw stays tight, breath slow through his nose, eyes fixed on the folded writ, shoulders settling into a resigned square while rain streaks crawl down the glass behind him. The bailiffs fall in at his elbows and walk him toward the side door, his slides scuffing stone in a steady rhythm, the jumpsuit hem swaying with each step. End mid-stride as the side door swings open onto a covered walkway, the writ-bearing bailiff half a step ahead, the motion clearly still in progress.",
"overall_soundscape: soft steady rain drumming on the courtroom's tall windows, the dry snap of the case band's clasp, paper folding, a low institutional hum under the hall's hush.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "intake_bay",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same mid-stride walk without resetting the motion: the bailiffs escort <Subject 1>, Case 6188, the same lean man with short-cropped dark brown hair, a fine pale scar over his left eyebrow, and narrow grey-green eyes, in his slate-grey jumpsuit, white socks, black slides, and black case band, from the covered walkway into the Meridian Center's intake bay. The camera keeps the incoming eye-level direction and reverses smoothly, pulling backward ahead of the group at their exact pace under the bay's ribbed roof. Photorealistic cool clinical daylight greys the wet concrete, the palette softening toward the center's pale pastels. The lead bailiff lifts an open grey tray and <Subject 1> sets his belt and a folding razor into it, watching the lid click shut; rain drips off the walkway edge behind them and puddle rings spread around their steps. At the far gate <Subject 2>, a calm stocky woman in a sand-beige uniform tunic and trousers, soft-soled white shoes, grey hair in a tight bun, ID lanyard, and cotton gloves, waits with one hand on the roller, her lips pressed in a neutral line. The gate grinds open and he steps through the frame mid-step, her gloved hand closing on his elbow. End mid-step as the gate frame slides past his shoulders, the camera still pulling backward at their pace.",
"overall_soundscape: the same steady rain continues on the intake bay's ribbed roof, boots splashing the wet walkway, a tray of effects rattling shut, gate rollers grinding slowly open.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "first_step",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward step without resetting the motion: his front foot lands inside the session wing as <Subject 2>, the same stocky attendant in her sand-beige uniform, soft-soled white shoes, grey bun, ID lanyard, and cotton gloves, takes his elbow and leads <Subject 1>, Case 6188, the lean dark-haired man with the pale scar over his left eyebrow and narrow grey-green eyes, in the slate-grey jumpsuit, white socks, and black case band, to the reclined chair in Session Room One. The camera eases up into a slow overhead crane angled down at the chair while the polished brass Step Ring lowers on its articulated arm. Photorealistic cool clinical daylight washes the room's pale panels, the slate-and-pastel palette at its coldest here. She straps his wrists and ankles with practiced presses, the chair hums back, and the ring pulses pale light down his body; across the take his adult frame slides continuously into its late teens, jaw and cheekbones softening, shoulders narrowing, hair thickening and growing shaggier over his ears, the jumpsuit seams straining and then hanging looser as he loses height against the headboard, his breathing quickening under the pulse. End with the ring light climbing toward a second pulse, his newly thin fingers curling on the armrests, the overhead camera still descending.",
"overall_soundscape: the same rain muffled through thick session glass, the room's clean system hum, and the Step Ring's deep pulse breathing every few seconds.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "teen_wing",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same settling of the ring's pale light without resetting the motion: <Subject 2>, the same attendant in sand-beige with cotton gloves and her grey bun, presses the armrests level and walks him out into the dusk-blue teen wing. <Subject 1>, Case 6188, the same face with its narrow grey-green eyes and the pale scar over the left eyebrow, moves on long lanky teen legs, dark brown hair thick and shaggy over his ears, wearing dusk-blue short-sleeve scrubs with matching drawstring trousers that finally fit, white socks, and a dusk-blue wristband on his left wrist. The camera drops to a low tracking shot at his hip height and follows him down the dorm corridor, doorways and height-marked doorframes sliding past in the cool clinical daylight, the slate-and-pastel palette warmed by the wing's blue panels. His stride keeps an adult's rhythm inside the teen frame, hands opening and closing at his sides, jaw set, eyes flicking to the pencil ticks climbing each doorframe; rain traces the window at the corridor's end. He stops at the nearest doorframe, reads the fresh marks, and reaches up. End mid-reach, fingertips a breath from the lowest fresh mark, the low camera still tracking in.",
"overall_soundscape: the same low system hum continues under socked steps on linoleum, rain tracing the dorm windows, a pencil line scratching fresh onto the doorframe.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "second_step",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same upward reach without resetting the motion: his fingertips hover at the pencil mark as <Subject 2>, the same attendant in sand-beige with cotton gloves, closes a hand on his shoulder and turns him back toward Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under thick shaggy dark brown hair, shuffles on lanky teen legs in the dusk-blue scrubs, white socks, and dusk-blue wristband, back to the reclined chair. The same low tracking camera follows him in and rises gently into the overhead angle as the chair reclines and the brass Step Ring descends for the second pulse. Photorealistic cool clinical daylight, the room's slate-and-pastel palette unchanged, rain muffled on the thick glass. The pulse runs and his teen frame slides continuously down through preteen across the take, length draining out of his legs, hands shrinking inside the sleeves, the scrubs swallowing him, the collar sliding off one shoulder, his hair settling into a soft bowl cut over his ears, breath hitching once and steadying. End with the ring still pulsing pale above him, his small hands lost inside the cuffs, the overhead camera holding.",
"overall_soundscape: the same muffled rain and system hum return with the Step Ring's deep pulse, cuff fabric sliding over small hands, the chair's motor easing the backrest down.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "courtyard_drills",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same fading of the ring's pulse without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, presses the chair level, helps him down, and walks him out to the walled therapy courtyard. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow, dark brown hair in a soft bowl cut, stands small at preteen height in the sage tunic with soft grey trousers, white socks, and a sage wristband. The overhead camera eases down to his eye level and tracks him sideways across the courtyard's soft rubber flooring under a canopy that sheds the steady rain. Photorealistic cool clinical daylight warms into the wing's sage and sand tones within the slate-and-pastel palette. She sets wooden blocks on a low bench and guides him through slow balance walks and a stacking row, her gloved hands correcting his elbows; he works with an adult's concentration inside the child's frame, tongue pressed to his lip, matching each block to its neighbor. End as he leans mid-reach for the next block, the tunic hem lifting, her hand hovering behind his shoulder, the sideways track holding.",
"overall_soundscape: the same rain softens on the courtyard canopy, wooden blocks clicking onto the bench, the attendant's calm steps circling, a distant pneumatic door breathing.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "third_step",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward lean of his reach without resetting the motion: <Subject 2>'s gloved hand intercepts the block, sets it back on the bench, and turns him inside toward Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under the soft dark bowl cut, walks at preteen height in the sage tunic, soft grey trousers, white socks, and sage wristband. The sideways tracking camera follows them through the doorway and settles into a slow push-in on the reclined chair as the brass Step Ring descends for the third pulse. Photorealistic cool clinical daylight returns in the room's coldest slate-and-pastel wash, rain steady on the thick glass. The pulse runs and his preteen frame settles continuously into a small child across the take, the tunic spreading into a tent around him, trousers pooling at his ankles, wrists thinning inside the sleeves, his features rounding, hair fine and straight over his brow, legs swinging where they no longer reach the floor. End with the ring light climbing again, socked heels together on the seat's edge, the push-in still creeping.",
"overall_soundscape: the same rain and hum carry back inside from the courtyard, the Step Ring's pulse deepening again, chair motor whirring as the backrest reclines.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "fitting_room",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same climbing of the ring light without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, eases the chair level, takes his hand, and leads him into the fitting room. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow, fine dark bowl-cut hair, stands at small child height and climbs onto a low wooden stool in white socks. The camera holds a slow quarter-orbit around the stool at his eye level, the wing's butter-yellow panels warming the slate-and-pastel palette, photorealistic cool daylight softened by the wide mirror. She buttons him into a butter-yellow child smock that ties at the back, smoothing the shoulders down over his frame, then folds the wing's standard care padding and folded supplies into the open cubby beside the mirror, stacking each shelf in even rows; he watches his mirrored reflection being dressed, fingers curling on the stool's edge, weight shifting from foot to foot. End as she draws the smock's back strap tight and smooths it flat, the orbit still holding at his eye level.",
"overall_soundscape: the same rain continues on the wing glass, folded supplies sliding into the cubby, buttons popping through cloth, small socked feet shifting on the wooden stool.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "fourth_step",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same tightening of the smock's back strap without resetting the motion: <Subject 2>, the same sand-beige attendant, pats the strap flat, takes his hand, and walks him back through the wing to Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow, fine dark bowl-cut hair, toddles at small child height in the butter-yellow smock, white socks, and yellow wristband. The orbit completes and the camera falls in at his shoulder, following him down the wing and rising into the overhead angle as the chair reclines and the brass Step Ring lowers for the fourth pulse. Photorealistic cool clinical daylight, slate-and-pastel palette, rain unbroken on the thick glass. The pulse sinks his small child frame continuously into a young child across the take, the smock tenting wider, limbs shortening, cheeks filling, hair softening into a fine fringe; she slips the smock over his head and draws a peach short-sleeve play set and matching shorts onto him, the cuffs swallowing his wrists, and settles him back against the chair. End with his socked toes settling onto the extended footrest, the peach wristband catching the pale light, the overhead camera holding.",
"overall_soundscape: the same rain and hum continue with the Step Ring's fourth deep pulse, soft fabric slipping over his head, the chair motor easing back under him.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "playroom_circle",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same settling of his toes onto the footrest without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, lifts him down from the chair and walks him into the young-child wing's playroom. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under the fine dark fringe, stands at young-child height in the peach short-sleeve play set and shorts, white socks, and peach wristband. The camera settles low at seated height and creeps laterally along the carpet circle, photorealistic cool clinical daylight warmed by the playroom's peach and cream tones, rain pattering the long glass wall. Two other cases work puzzles across the circle, a girl in a sage tunic and a boy in a yellow smock, both silent. She sets a wooden puzzle board in front of him; he studies it with an adult's furrow in a round young face, fits two pieces, glances up at her, and returns to the board while a soft block rolls loose across the carpet. End as he tips forward mid-reach after it, one palm already down on the carpet, the lateral creep still following.",
"overall_soundscape: the same rain patters on the playroom glass, soft blocks thudding on carpet, small palms slapping the floor, the wing's quiet hum underneath.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "final_step",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward tip of his reach without resetting the motion: <Subject 2>'s gloved hand catches the soft block first, steadies his shoulder, and lifts him up from the carpet circle to carry him back through the wing to Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under the fine dark fringe, rides her arm at young-child height in the peach play set, white socks, and peach wristband. The lateral creep camera rises with her lift and holds a slow push-in as she settles him into the reclined chair and the brass Step Ring descends for the fifth and final pulse. Photorealistic cool clinical daylight, the room's slate-and-pastel palette at its quietest, rain steady on the glass. The pulse runs and his young-child frame dissolves continuously into a toddler across the take, limbs rounding and shortening, baby fat filling his cheeks, the fringe thinning into fine soft hair over a round brow; she slips the play set off and dresses him in a cream footed sleeper with grip soles, pressing each snap closed with practiced hands. End mid-motion as she lifts him from the chair toward the waiting crib, his cream-sleepered frame curling against her shoulder, the push-in still creeping.",
"overall_soundscape: the same rain and hum continue with the Step Ring's final deep pulse, sleeper snaps pressing closed, then the crib mobile's slow chimes turning overhead.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
},
{
"id": "nursery_morning",
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same mid-motion lift without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, lowers <Subject 1>, Case 6188, a toddler of barely 0.8 meters with fine soft dark hair over a round face still holding the narrow grey-green eyes and the pale scar over the left eyebrow, dressed in the cream footed sleeper with grip soles and a cream wristband, gently into the cream crib. The mobile turns its slow wooden shapes overhead as the camera pushes in once and then retreats smoothly, the clinical night light easing continuously into warm morning light while the rain thins to a drizzle on the district glass, and the camera glides back through the observation window and comes to rest outside. Photorealistic morning warmth settles over the nursery's cream-and-pastel palette. She completes the quiet changing-table routine with practiced, unhurried presses, carries him to the playmat, and settles cross-legged beside him; he sits, sways, catches his balance, and grasps the soft block with both hands, turning it over. End at rest on the window-framed view: the toddler working the block on the playmat, the attendant beside him, the mobile's chimes continuing over the still morning room.",
"overall_soundscape: the same rain thins to a drizzle on the nursery glass, soft morning hum, the mobile's slow chimes continuing, a soft block thudding once on the playmat.",
"non_diegetic_music: N/A"
],
"duration_seconds": 10
}
]
}

View File

@ -1,14 +0,0 @@
{
"id": "sentencing_hall",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Opening scene. <Subject 1>, Case 6188, a lean adult man about 1.8 meters tall with short-cropped dark brown hair, a fine pale scar over his left eyebrow, and narrow grey-green eyes, stands at the rail of a function courtroom in a slate-grey jumpsuit, white socks in plain black slides, and a fresh black case band sealed around his left wrist. The camera starts low behind the courtroom rail and pushes in slowly at eye level, holding him and the two bailiffs in one stable line. Photorealistic, cool clinical daylight falls through tall windows streaked with rain, laying soft grey light across the dark wood rail and the hall's slate-and-pastel palette. The nearer bailiff presses the case band clasp closed with a snap and the second folds the writ of Functional Reassignment into his jacket; <Subject 1>'s jaw stays tight, breath slow through his nose, eyes fixed on the folded writ, shoulders settling into a resigned square while rain streaks crawl down the glass behind him. The bailiffs fall in at his elbows and walk him toward the side door, his slides scuffing stone in a steady rhythm, the jumpsuit hem swaying with each step. End mid-stride as the side door swings open onto a covered walkway, the writ-bearing bailiff half a step ahead, the motion clearly still in progress.",
"overall_soundscape: soft steady rain drumming on the courtroom's tall windows, the dry snap of the case band's clasp, paper folding, a low institutional hum under the hall's hush.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "Case 6188 is mid-stride through the courtroom's open side door onto the covered walkway, the writ-bearing bailiff half a step ahead and the second bailiff at his elbow, rain falling past the roof edge. The camera is still pushing in at eye level behind them, the black case band catching the grey light.",
"soundscape": "overall_soundscape: soft steady rain drumming on the courtroom's tall windows, the dry snap of the case band's clasp, paper folding, a low institutional hum under the hall's hush.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "intake_bay",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same mid-stride walk without resetting the motion: the bailiffs escort <Subject 1>, Case 6188, the same lean man with short-cropped dark brown hair, a fine pale scar over his left eyebrow, and narrow grey-green eyes, in his slate-grey jumpsuit, white socks, black slides, and black case band, from the covered walkway into the Meridian Center's intake bay. The camera keeps the incoming eye-level direction and reverses smoothly, pulling backward ahead of the group at their exact pace under the bay's ribbed roof. Photorealistic cool clinical daylight greys the wet concrete, the palette softening toward the center's pale pastels. The lead bailiff lifts an open grey tray and <Subject 1> sets his belt and a folding razor into it, watching the lid click shut; rain drips off the walkway edge behind them and puddle rings spread around their steps. At the far gate <Subject 2>, a calm stocky woman in a sand-beige uniform tunic and trousers, soft-soled white shoes, grey hair in a tight bun, ID lanyard, and cotton gloves, waits with one hand on the roller, her lips pressed in a neutral line. The gate grinds open and he steps through the frame mid-step, her gloved hand closing on his elbow. End mid-step as the gate frame slides past his shoulders, the camera still pulling backward at their pace.",
"overall_soundscape: the same steady rain continues on the intake bay's ribbed roof, boots splashing the wet walkway, a tray of effects rattling shut, gate rollers grinding slowly open.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "He is stepping through the open intake gate frame mid-step, the attendant's gloved hand closing on his elbow, the sealed grey tray carried away behind him. The camera still pulls backward ahead of the group at their pace, rain drumming on the ribbed roof.",
"soundscape": "overall_soundscape: the same steady rain continues on the intake bay's ribbed roof, boots splashing the wet walkway, a tray of effects rattling shut, gate rollers grinding slowly open.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "first_step",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward step without resetting the motion: his front foot lands inside the session wing as <Subject 2>, the same stocky attendant in her sand-beige uniform, soft-soled white shoes, grey bun, ID lanyard, and cotton gloves, takes his elbow and leads <Subject 1>, Case 6188, the lean dark-haired man with the pale scar over his left eyebrow and narrow grey-green eyes, in the slate-grey jumpsuit, white socks, and black case band, to the reclined chair in Session Room One. The camera eases up into a slow overhead crane angled down at the chair while the polished brass Step Ring lowers on its articulated arm. Photorealistic cool clinical daylight washes the room's pale panels, the slate-and-pastel palette at its coldest here. She straps his wrists and ankles with practiced presses, the chair hums back, and the ring pulses pale light down his body; across the take his adult frame slides continuously into its late teens, jaw and cheekbones softening, shoulders narrowing, hair thickening and growing shaggier over his ears, the jumpsuit seams straining and then hanging looser as he loses height against the headboard, his breathing quickening under the pulse. End with the ring light climbing toward a second pulse, his newly thin fingers curling on the armrests, the overhead camera still descending.",
"overall_soundscape: the same rain muffled through thick session glass, the room's clean system hum, and the Step Ring's deep pulse breathing every few seconds.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "The chair holds his late-teens frame reclined, jaw softened, hair thicker and shaggier, jumpsuit cuffs riding up his thin wrists, fingers curling on the armrests. The Step Ring's light climbs toward a second pulse and the overhead camera is still descending.",
"soundscape": "overall_soundscape: the same rain muffled through thick session glass, the room's clean system hum, and the Step Ring's deep pulse breathing every few seconds.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "teen_wing",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same settling of the ring's pale light without resetting the motion: <Subject 2>, the same attendant in sand-beige with cotton gloves and her grey bun, presses the armrests level and walks him out into the dusk-blue teen wing. <Subject 1>, Case 6188, the same face with its narrow grey-green eyes and the pale scar over the left eyebrow, moves on long lanky teen legs, dark brown hair thick and shaggy over his ears, wearing dusk-blue short-sleeve scrubs with matching drawstring trousers that finally fit, white socks, and a dusk-blue wristband on his left wrist. The camera drops to a low tracking shot at his hip height and follows him down the dorm corridor, doorways and height-marked doorframes sliding past in the cool clinical daylight, the slate-and-pastel palette warmed by the wing's blue panels. His stride keeps an adult's rhythm inside the teen frame, hands opening and closing at his sides, jaw set, eyes flicking to the pencil ticks climbing each doorframe; rain traces the window at the corridor's end. He stops at the nearest doorframe, reads the fresh marks, and reaches up. End mid-reach, fingertips a breath from the lowest fresh mark, the low camera still tracking in.",
"overall_soundscape: the same low system hum continues under socked steps on linoleum, rain tracing the dorm windows, a pencil line scratching fresh onto the doorframe.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "He stands at the height-mark doorframe in the dusk-blue corridor, arm raised, fingertips a breath from the lowest fresh pencil mark, weight rocking up onto his toes. The low camera is still tracking in at his hip height.",
"soundscape": "overall_soundscape: the same low system hum continues under socked steps on linoleum, rain tracing the dorm windows, a pencil line scratching fresh onto the doorframe.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "second_step",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same upward reach without resetting the motion: his fingertips hover at the pencil mark as <Subject 2>, the same attendant in sand-beige with cotton gloves, closes a hand on his shoulder and turns him back toward Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under thick shaggy dark brown hair, shuffles on lanky teen legs in the dusk-blue scrubs, white socks, and dusk-blue wristband, back to the reclined chair. The same low tracking camera follows him in and rises gently into the overhead angle as the chair reclines and the brass Step Ring descends for the second pulse. Photorealistic cool clinical daylight, the room's slate-and-pastel palette unchanged, rain muffled on the thick glass. The pulse runs and his teen frame slides continuously down through preteen across the take, length draining out of his legs, hands shrinking inside the sleeves, the scrubs swallowing him, the collar sliding off one shoulder, his hair settling into a soft bowl cut over his ears, breath hitching once and steadying. End with the ring still pulsing pale above him, his small hands lost inside the cuffs, the overhead camera holding.",
"overall_soundscape: the same muffled rain and system hum return with the Step Ring's deep pulse, cuff fabric sliding over small hands, the chair's motor easing the backrest down.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "The chair holds his small preteen frame half-swallowed by the dusk-blue scrubs, hands lost inside the cuffs, the Step Ring still pulsing pale above him. The overhead camera holds steady, rain muffled on the glass.",
"soundscape": "overall_soundscape: the same muffled rain and system hum return with the Step Ring's deep pulse, cuff fabric sliding over small hands, the chair's motor easing the backrest down.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "courtyard_drills",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same fading of the ring's pulse without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, presses the chair level, helps him down, and walks him out to the walled therapy courtyard. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow, dark brown hair in a soft bowl cut, stands small at preteen height in the sage tunic with soft grey trousers, white socks, and a sage wristband. The overhead camera eases down to his eye level and tracks him sideways across the courtyard's soft rubber flooring under a canopy that sheds the steady rain. Photorealistic cool clinical daylight warms into the wing's sage and sand tones within the slate-and-pastel palette. She sets wooden blocks on a low bench and guides him through slow balance walks and a stacking row, her gloved hands correcting his elbows; he works with an adult's concentration inside the child's frame, tongue pressed to his lip, matching each block to its neighbor. End as he leans mid-reach for the next block, the tunic hem lifting, her hand hovering behind his shoulder, the sideways track holding.",
"overall_soundscape: the same rain softens on the courtyard canopy, wooden blocks clicking onto the bench, the attendant's calm steps circling, a distant pneumatic door breathing.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "He leans forward mid-reach for the next wooden block on the courtyard bench, the sage tunic hem lifting, the attendant's gloved hand hovering behind his shoulder. The camera holds its sideways track at his eye level.",
"soundscape": "overall_soundscape: the same rain softens on the courtyard canopy, wooden blocks clicking onto the bench, the attendant's calm steps circling, a distant pneumatic door breathing.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "third_step",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward lean of his reach without resetting the motion: <Subject 2>'s gloved hand intercepts the block, sets it back on the bench, and turns him inside toward Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under the soft dark bowl cut, walks at preteen height in the sage tunic, soft grey trousers, white socks, and sage wristband. The sideways tracking camera follows them through the doorway and settles into a slow push-in on the reclined chair as the brass Step Ring descends for the third pulse. Photorealistic cool clinical daylight returns in the room's coldest slate-and-pastel wash, rain steady on the thick glass. The pulse runs and his preteen frame settles continuously into a small child across the take, the tunic spreading into a tent around him, trousers pooling at his ankles, wrists thinning inside the sleeves, his features rounding, hair fine and straight over his brow, legs swinging where they no longer reach the floor. End with the ring light climbing again, socked heels together on the seat's edge, the push-in still creeping.",
"overall_soundscape: the same rain and hum carry back inside from the courtyard, the Step Ring's pulse deepening again, chair motor whirring as the backrest reclines.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "The chair holds his small child frame with the sage tunic spread into a tent around him, socked heels together on the seat's edge. The Step Ring's light climbs again and the slow push-in is still creeping.",
"soundscape": "overall_soundscape: the same rain and hum carry back inside from the courtyard, the Step Ring's pulse deepening again, chair motor whirring as the backrest reclines.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "fitting_room",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same climbing of the ring light without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, eases the chair level, takes his hand, and leads him into the fitting room. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow, fine dark bowl-cut hair, stands at small child height and climbs onto a low wooden stool in white socks. The camera holds a slow quarter-orbit around the stool at his eye level, the wing's butter-yellow panels warming the slate-and-pastel palette, photorealistic cool daylight softened by the wide mirror. She buttons him into a butter-yellow child smock that ties at the back, smoothing the shoulders down over his frame, then folds the wing's standard care padding and folded supplies into the open cubby beside the mirror, stacking each shelf in even rows; he watches his mirrored reflection being dressed, fingers curling on the stool's edge, weight shifting from foot to foot. End as she draws the smock's back strap tight and smooths it flat, the orbit still holding at his eye level.",
"overall_soundscape: the same rain continues on the wing glass, folded supplies sliding into the cubby, buttons popping through cloth, small socked feet shifting on the wooden stool.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "The attendant draws the smock's back strap tight behind him on the stool, one hand flattening the butter-yellow fabric across his shoulders, the cubby door still open beside the mirror. The quarter-orbit camera holds at his eye level.",
"soundscape": "overall_soundscape: the same rain continues on the wing glass, folded supplies sliding into the cubby, buttons popping through cloth, small socked feet shifting on the wooden stool.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "fourth_step",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same tightening of the smock's back strap without resetting the motion: <Subject 2>, the same sand-beige attendant, pats the strap flat, takes his hand, and walks him back through the wing to Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow, fine dark bowl-cut hair, toddles at small child height in the butter-yellow smock, white socks, and yellow wristband. The orbit completes and the camera falls in at his shoulder, following him down the wing and rising into the overhead angle as the chair reclines and the brass Step Ring lowers for the fourth pulse. Photorealistic cool clinical daylight, slate-and-pastel palette, rain unbroken on the thick glass. The pulse sinks his small child frame continuously into a young child across the take, the smock tenting wider, limbs shortening, cheeks filling, hair softening into a fine fringe; she slips the smock over his head and draws a peach short-sleeve play set and matching shorts onto him, the cuffs swallowing his wrists, and settles him back against the chair. End with his socked toes settling onto the extended footrest, the peach wristband catching the pale light, the overhead camera holding.",
"overall_soundscape: the same rain and hum continue with the Step Ring's fourth deep pulse, soft fabric slipping over his head, the chair motor easing back under him.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "He sits back in the reclined chair in the peach play set and shorts, socked toes settling onto the extended footrest, the folded smock draped over the attendant's arm. The overhead camera holds, rain unbroken on the glass.",
"soundscape": "overall_soundscape: the same rain and hum continue with the Step Ring's fourth deep pulse, soft fabric slipping over his head, the chair motor easing back under him.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "playroom_circle",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same settling of his toes onto the footrest without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, lifts him down from the chair and walks him into the young-child wing's playroom. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under the fine dark fringe, stands at young-child height in the peach short-sleeve play set and shorts, white socks, and peach wristband. The camera settles low at seated height and creeps laterally along the carpet circle, photorealistic cool clinical daylight warmed by the playroom's peach and cream tones, rain pattering the long glass wall. Two other cases work puzzles across the circle, a girl in a sage tunic and a boy in a yellow smock, both silent. She sets a wooden puzzle board in front of him; he studies it with an adult's furrow in a round young face, fits two pieces, glances up at her, and returns to the board while a soft block rolls loose across the carpet. End as he tips forward mid-reach after it, one palm already down on the carpet, the lateral creep still following.",
"overall_soundscape: the same rain patters on the playroom glass, soft blocks thudding on carpet, small palms slapping the floor, the wing's quiet hum underneath.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "He tips forward mid-reach after the rolling soft block on the carpet circle, one palm already down, the other arm reaching, the two other cases still working their puzzles beside him. The low camera creeps laterally with him.",
"soundscape": "overall_soundscape: the same rain patters on the playroom glass, soft blocks thudding on carpet, small palms slapping the floor, the wing's quiet hum underneath.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "final_step",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward tip of his reach without resetting the motion: <Subject 2>'s gloved hand catches the soft block first, steadies his shoulder, and lifts him up from the carpet circle to carry him back through the wing to Session Room One. <Subject 1>, Case 6188, the same narrow grey-green eyes and pale scar over the left eyebrow under the fine dark fringe, rides her arm at young-child height in the peach play set, white socks, and peach wristband. The lateral creep camera rises with her lift and holds a slow push-in as she settles him into the reclined chair and the brass Step Ring descends for the fifth and final pulse. Photorealistic cool clinical daylight, the room's slate-and-pastel palette at its quietest, rain steady on the glass. The pulse runs and his young-child frame dissolves continuously into a toddler across the take, limbs rounding and shortening, baby fat filling his cheeks, the fringe thinning into fine soft hair over a round brow; she slips the play set off and dresses him in a cream footed sleeper with grip soles, pressing each snap closed with practiced hands. End mid-motion as she lifts him from the chair toward the waiting crib, his cream-sleepered frame curling against her shoulder, the push-in still creeping.",
"overall_soundscape: the same rain and hum continue with the Step Ring's final deep pulse, sleeper snaps pressing closed, then the crib mobile's slow chimes turning overhead.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "The attendant lifts him from the chair in mid-motion, his cream-sleepered toddler frame curling against her shoulder, the Step Ring's light fading behind them. The camera pushes slowly in as she turns toward the waiting crib.",
"soundscape": "overall_soundscape: the same rain and hum continue with the Step Ring's final deep pulse, sleeper snaps pressing closed, then the crib mobile's slow chimes turning overhead.",
"dialogue_used": []
}
}

View File

@ -1,14 +0,0 @@
{
"id": "nursery_morning",
"duration_seconds": 10,
"prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same mid-motion lift without resetting the motion: <Subject 2>, the same sand-beige attendant with cotton gloves and grey bun, lowers <Subject 1>, Case 6188, a toddler of barely 0.8 meters with fine soft dark hair over a round face still holding the narrow grey-green eyes and the pale scar over the left eyebrow, dressed in the cream footed sleeper with grip soles and a cream wristband, gently into the cream crib. The mobile turns its slow wooden shapes overhead as the camera pushes in once and then retreats smoothly, the clinical night light easing continuously into warm morning light while the rain thins to a drizzle on the district glass, and the camera glides back through the observation window and comes to rest outside. Photorealistic morning warmth settles over the nursery's cream-and-pastel palette. She completes the quiet changing-table routine with practiced, unhurried presses, carries him to the playmat, and settles cross-legged beside him; he sits, sways, catches his balance, and grasps the soft block with both hands, turning it over. End at rest on the window-framed view: the toddler working the block on the playmat, the attendant beside him, the mobile's chimes continuing over the still morning room.",
"overall_soundscape: the same rain thins to a drizzle on the nursery glass, soft morning hum, the mobile's slow chimes continuing, a soft block thudding once on the playmat.",
"non_diegetic_music: N/A"
],
"_handoff": {
"end_state": "The camera has pulled back through the observation window and comes to rest: beyond the glass the toddler works the soft block on the playmat in warm morning light, the attendant seated cross-legged beside him. Rain thins on the district glass and the mobile's chimes continue over the still room.",
"soundscape": "overall_soundscape: the same rain thins to a drizzle on the nursery glass, soft morning hum, the mobile's slow chimes continuing, a soft block thudding once on the playmat.",
"dialogue_used": []
}
}

View File

@ -38,7 +38,7 @@ never compose a scene from a guide the user has not approved.
``` ```
Expect the suite to pass and the second command to print Expect the suite to pass and the second command to print
`PLAN OK` with `total delivered: 1042 frames = 43.42s`. `PLAN OK` with `total delivered: 685 frames = 28.54s`.
## Step 1 — Build the scene guide (outline) ## Step 1 — Build the scene guide (outline)
@ -78,6 +78,22 @@ voice: <voice text>&` — copy the identity and voice text verbatim into that su
inventory string (the & marks are delimiters, never content); the scene turns must carry inventory string (the & marks are delimiters, never content); the scene turns must carry
them into subject_definitions untouched. them into subject_definitions untouched.
### Subject pinning (three forms — decide here, use in every scene)
Each `subjects` string decides how that subject is pinned in the final prompt, and every
scene's `subject_definitions` line must use exactly that form:
- **No reference image** — `<Subject N> is a <age> ... with <hair>, <eyes> and <build>`:
physical traits only (age, hair, eyes, build, skin tone). Never clothing, never a
nude/bare skin state — the scene body carries the wardrobe.
- **Reference image** — `<Subject N> binds to <Image N> : <what the image actually
shows>`: a faithful description of the image; the writer is multimodal.
- **User refmod** — the user's story carries their own tested character block, like
`&identity: ...; voice: ...&`: strip the `&` wrapper and carry the text verbatim into
`<Subject N> identity: ...; voice: ...`. Exempt from the no-image content bans.
No form carries a length cap — describe fully what the model needs.
## Step 2 — Present the guide for approval (do not skip) ## Step 2 — Present the guide for approval (do not skip)
Show the user the finished guide in plain language — one line per scene: number, id, Show the user the finished guide in plain language — one line per scene: number, id,
@ -108,24 +124,27 @@ and never summarize.
(`--total` is the scene count from the outline; `--prev` chains the previous file.) (`--total` is the scene count from the outline; `--prev` chains the previous file.)
Each fragment is exactly this shape: Each fragment is exactly this shape (the six-section Ref2VA prompt; the subject lines
use the pinning form chosen in the guide):
```json ```json
{ {
"id": "from the outline row", "id": "from the outline row",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"subject_definitions:", "subject_definitions:",
"<Subject 1> ...one of the three definition forms (see craft rules below)...", "<Subject 1> identity: ...; voice: ...",
"<Subject 2> is a ...",
"summary:", "summary:",
"[reference generation]", "[reference generation]",
"retention_analysis:", "retention_analysis:",
"<Subject 1>: fully_preserved", "<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:", "detailed_description:",
"...1-2 style sentences from the guide's style...", "<1-2 style sentences>",
"[Shot 1] ...the full scene composition...", "[Shot 1] ...",
"overall_soundscape:", "overall_soundscape:",
"...the sound the scene itself makes...", "<the scene's own sound>",
"non_diegetic_music:", "non_diegetic_music:",
"N/A" "N/A"
], ],
@ -155,7 +174,15 @@ system prompt emits. Report the delivered frames/seconds it prints.
## Step 5 — Deliver (the node makes the multi-clip video) ## Step 5 — Deliver (the node makes the multi-clip video)
The user pastes the full contents of `plan.json` into the Plan node's `plan_json_input` The user pastes the full contents of `plan.json` into the Plan node's `plan_json_input`
and queues ONCE. The node renders each shot as its own clip and chains the clips through and queues ONCE. The `plan_json_input` field must contain EXACTLY the one JSON object:
first character `{`, last character `}`, nothing before or after. If ComfyUI rejects the
paste with `H3 Chain Plan JSON is invalid: Extra data: line N column M`, the field holds
more than the single JSON object — leftover text from an earlier plan ahead of the paste,
or a duplicated fragment of it. Clear the field completely (select all, delete), paste
once, and check that the field starts with `{` and ends with `}`. If `plan_json_input`
is wired to another node, the paste belongs in that source node's field instead.
The node renders each shot as its own clip and chains the clips through
the H3 context loop — scene 2 continues from scene 1's final frames, scene 3 from scene the H3 context loop — scene 2 continues from scene 1's final frames, scene 3 from scene
2's — producing one continuous multi-clip video. Every scene after the first loses about 2's — producing one continuous multi-clip video. Every scene after the first loses about
1 second to the fixed overlap trim that stitches the clips together. 1 second to the fixed overlap trim that stitches the clips together.
@ -168,37 +195,37 @@ and sampler latent or the video can come out square.
- Raw JSON only: first char `{`, last char `}`, no markdown fences, no prose, double - Raw JSON only: first char `{`, last char `}`, no markdown fences, no prose, double
quotes, no trailing commas, no raw line break inside a string. quotes, no trailing commas, no raw line break inside a string.
- The `prompt` array carries the six native H3 ref2va sections in canonical order — - The `prompt` array compiles to the six-section Ref2VA prompt in canonical order —
`subject_definitions:` / `summary:` / `retention_analysis:` / `detailed_description:` / `subject_definitions:` / `summary:` / `retention_analysis:` / `detailed_description:` /
`overall_soundscape:` / `non_diegetic_music:` — one line per element, headers on their `overall_soundscape:` / `non_diegetic_music:` — with `summary:` exactly
own line; `summary` is exactly `[reference generation]`; every scene is its own single `[reference generation]`, one `<Subject N>: fully_preserved` retention line per defined
H3 generation, so the bracket number is always `[Shot 1]`. subject, 1–2 style sentences before the body, and `[Shot 1]` starting the body (every
- The full look is re-stated in EVERY scene, and it lives in the [Shot 1] body — scene is its own single H3 generation, so the bracket number is always 1).
subject_definitions only anchors identity, in exactly one of three forms per line: - `subject_definitions` lines come in exactly one of three forms: `<Subject N> is a ...`
a character with NO reference image gets physical traits only ("<Subject 1> is a (no image — physical traits only, never clothing, never a nude/bare skin state),
20 year old woman with blonde hair, blue eyes and a medium build" — age, hair, `<Subject N> binds to <Image N> : ...` (reference image — faithful to the image), or
eyes, build; never clothing, never a nude/bare state); a character WITH a `<Subject N> identity: ...; voice: ...` (user refmod — the story's tested wording
reference image gets "<Subject 1> binds to <Image 1> : what the image actually verbatim, the `&...&` wrapper stripped; exempt from the no-image bans). No form carries
shows" (the writing LLM is multimodal — look at it); and a user a length cap.
refmod (`sam:&identity: ...; voice: ...&` in the story's characters block) is - **Refmod isolation:** a refmod subject is referenced by the bare `<Subject N>` token
carried through VERBATIM as "<Subject 1> identity: ...; voice: ..." — never everywhere outside `subject_definitions` — the body never re-states the subject's name
rewritten, never trimmed, exempt from every defs limit; the refmod text lives and never re-describes the look or voice the refmod already pins (the workflow injects
ONLY in subject_definitions, and the [Shot 1] body references that subject by the reference blocks itself; repeating them fights the injection). No-image and binds
its `<Subject N>` label alone — never restating the refmod's identity, voice, subjects keep the opposite treatment: their full look is still woven into every scene
or look. For a no-image or binds subject, weave the head-to-toe body.
look into the action as the scene exposes it — hair, build, and face as they - No-image and binds subjects are self-sufficient: re-state `<Subject 1>` and their full
move, each garment where the action touches it, footwear where steps land — look in EVERY scene — identity, hair, build, every garment head to toe, footwear, and
nothing is shared between scenes or from a prefix. Name every body part the the visual style — nothing is shared between scenes or from a prefix. Name every body
clothing leaves uncovered and call uncovered skin "nude", never "bare", where the part the clothing leaves uncovered and call uncovered skin "nude", never "bare"
scene shows it ("a man wearing a white tank-top with nude arms"). ("a man wearing a white tank-top with nude arms").
- Scene composition, not word count: the [Shot 1] body must carry every element — - Scene composition, not word count: the body must carry every element —
subject anchor (`<Subject 1>` referenced; the full look woven in — a refmod subject needs only its label), style anchor, subject anchor (`<Subject 1>` restated; for no-image/binds subjects with wardrobe),
continuity anchor style anchor, continuity anchor
(opens inside the previous end_state), blocking, one camera idea, performance beats, (opens inside the previous end_state), blocking, one camera idea, performance beats,
physical texture (wardrobe behavior, weather on skin and fabric, lighting), and a physical texture (wardrobe behavior, weather on skin and fabric, lighting), and a
concrete mid-action exit state. Never pad to a length and never hand in a thin concrete mid-action exit state. Never pad to a length and never hand in a thin
one-liner — a scene with every element is long enough on its own. The validator only one-liner — a scene with every element is long enough on its own. The validator only
trips a thin-scene floor (about 100 words); there is no target band. trips a thin-scene floor (about 100 body words); there is no target band.
- Decompose every action into body mechanics: never a single outcome verb ("picks up - Decompose every action into body mechanics: never a single outcome verb ("picks up
the ball", "crosses her legs") — write the ordered joint-level sequence ("bends his the ball", "crosses her legs") — write the ordered joint-level sequence ("bends his
knees, grabs the ball with his left hand, then stands back up"; "raises her left knees, grabs the ball with his left hand, then stands back up"; "raises her left
@ -240,8 +267,8 @@ and sampler latent or the video can come out square.
a "silently with mouth {position}" cue before and after every span. Mirror each span in a "silently with mouth {position}" cue before and after every span. Mirror each span in
`_handoff.dialogue_used` (words only, no marker); empty array when silent. `_handoff.dialogue_used` (words only, no marker); empty array when silent.
- `_handoff` must be truthful: `end_state` describes the real final frame (2–3 sentences: - `_handoff` must be truthful: `end_state` describes the real final frame (2–3 sentences:
pose, motion in progress, camera state); `soundscape` copies the soundscape content pose, motion in progress, camera state); `soundscape` copies the overall_soundscape text
line (the element after the `overall_soundscape:` header) character for character. (no section label) character for character.
- Keys are exactly `id`, `duration_seconds`, `prompt`, `_handoff`; add `steps` or `seed` - Keys are exactly `id`, `duration_seconds`, `prompt`, `_handoff`; add `steps` or `seed`
only when the user explicitly asked. Every duration 2–15 s. only when the user explicitly asked. Every duration 2–15 s.

View File

@ -4,7 +4,8 @@
Stage 3 of the agent workflow (outline -> one scene per turn -> put it all together Stage 3 of the agent workflow (outline -> one scene per turn -> put it all together
at the end). Deterministic and byte-exact: fragment prompt strings are copied, never at the end). Deterministic and byte-exact: fragment prompt strings are copied, never
rewritten. Every fragment and the merged plan are validated with the same rules the rewritten. Every fragment and the merged plan are validated with the same rules the
node uses (replicated in verify_plan_rules.py) before anything is written. node uses (replicated in verify_plan_rules.py, including the six-section Ref2VA
prompt contract and refmod isolation) before anything is written.
The merged plan carries one shot per scene. The node renders each shot as its own The merged plan carries one shot per scene. The node renders each shot as its own
clip and chains the clips through the H3 context loop into one continuous video — clip and chains the clips through the H3 context loop into one continuous video —
@ -77,7 +78,7 @@ def main():
plan = load_json_file(args.check) plan = load_json_file(args.check)
print(f"Checking plan: {args.check}") print(f"Checking plan: {args.check}")
vpr.validate_plan(plan) vpr.validate_plan(plan)
vpr.check_h3_sections(plan) vpr.check_ref2va_sections(plan)
print("PLAN OK") print("PLAN OK")
return return
@ -131,7 +132,7 @@ def main():
print("\nMerged plan:") print("\nMerged plan:")
vpr.validate_plan(plan) vpr.validate_plan(plan)
vpr.check_h3_sections(plan) vpr.check_ref2va_sections(plan)
out_text = json.dumps(plan, indent=2, ensure_ascii=False) + "\n" out_text = json.dumps(plan, indent=2, ensure_ascii=False) + "\n"
if args.out: if args.out:

36
h3-plan-gate/h3_agent_workflow_examples/outline.json Executable file → Normal file
View File

@ -1,32 +1,38 @@
{ {
"title": "greenhouse_courier", "title": "candy_store",
"defaults": { "duration_seconds": 15, "steps": 20 }, "defaults": {
"duration_seconds": 10,
"steps": 5
},
"style_and_cast": { "style_and_cast": {
"subjects": ["<Subject 1> — an adult bicycle courier; mustard-yellow waterproof jacket, dark teal trousers, charcoal helmet, olive messenger bag"], "subjects": [
"style": "Photorealistic overcast rainy daylight, soft grey palette, stable landscape tracking shots at eye level", "<Subject 1> — Sam (user refmod, carry verbatim into subject_definitions): identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"soundscape_seed": "steady light rain drumming on glass roofing, low distant traffic hum" "<Subject 2> — the shopkeeper: 55 year old man, short grey hair, brown eyes, heavy build; green apron over a rolled-sleeve white shirt, dark trousers (look woven into each [Shot 1] body)"
],
"style": "Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain, static eye-level camera held at the child's height.",
"soundscape_seed": "the low hum of a candy-store fridge case, cellophane wrapper crinkle, muffled street traffic behind the front glass"
}, },
"scenes": [ "scenes": [
{ {
"index": 1, "index": 1,
"id": "greenhouse_arrival", "id": "candy_browse",
"duration_seconds": 15, "duration_seconds": 10,
"beat": "He wheels his silver bicycle through the greenhouse entrance out of the rain and crosses the interior threshold mid-stride.", "beat": "Sam browses the candy shelves, taps the glass in front of a swirl lollipop, and turns toward the counter mid-step.",
"dialogue": [] "dialogue": []
}, },
{ {
"index": 2, "index": 2,
"id": "greenhouse_aisle", "id": "lollipop_pickup",
"duration_seconds": 15, "duration_seconds": 10,
"beat": "He pushes the bike down the central aisle toward the potting table and slows it to a crawl, one hand reaching for the bag flap.", "beat": "Sam picks up the red-and-white swirl lollipop from the counter and his left hand finds the small red coin purse at his hip.",
"dialogue": [] "dialogue": []
}, },
{ {
"index": 3, "index": 3,
"id": "greenhouse_delivery", "id": "pay_shopkeeper",
"duration_seconds": 15, "duration_seconds": 10,
"beat": "He sets the kraft parcel on the wooden potting table and speaks his line as the delivery completes.", "beat": "Sam pays the shopkeeper, coins into the springing register drawer, lollipop raised high.",
"dialogue": ["Package for the greenhouse, right on time."] "dialogue": []
} }
] ]
} }

73
h3-plan-gate/h3_agent_workflow_examples/plan.json Executable file → Normal file
View File

@ -3,36 +3,75 @@
"" ""
], ],
"defaults": { "defaults": {
"duration_seconds": 15, "duration_seconds": 10,
"steps": 20 "steps": 5
}, },
"shots": [ "shots": [
{ {
"id": "greenhouse_arrival", "id": "candy_browse",
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Opening scene. <Subject 1>, the adult bicycle courier, steps off his silver bicycle at the wet gravel path outside the glass greenhouse and walks it slowly through the entrance, right hand on the handlebar, left thumb hooked under the strap of the olive messenger bag as the kraft paper parcel presses against his hip. The camera is a stable landscape tracking shot at eye level, gliding sideways at his exact walking pace, keeping his face, the parcel, and the fogged glass wall behind him in frame; the camera height and the direction of travel never change. Photorealistic overcast rainy daylight filters through the fogged panes and lays soft grey light across his mustard-yellow waterproof jacket; rain beads and long streaks run down the glass, and the wet paving glistens under the bicycle wheels. Droplets dot his charcoal helmet, his dark teal trousers darken at the shins, and each footstep squeezes a small splash between the gravel stones. He breathes steadily with a relaxed jaw, eyes flicking up to the aisle numbers as he passes benches of potted ferns and clay pots, and the kraft paper shifts faintly against the bag strap with every step. End while he is crossing the interior threshold, front wheel still rolling over the concrete sill, coat sleeve mid-swing, motion clearly still in progress.", "subject_definitions:",
"overall_soundscape: steady light rain drumming on glass roofing, wet tires crunching over wet paving, low distant traffic hum under a grey sky.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] <Subject 1> stands rooted in front of the counter-height candy shelves, both small hands hovering an inch from the glass. The camera holds a static low wide angle at his exact eye height, the ranked candy jars and bright wrapper rows sharp in shallow focus around him. <Subject 2>, the heavy-set shopkeeper with short grey hair in his green apron over a rolled-sleeve white shirt, sets a fresh jar of lollipops on the counter's far end behind the register, his big hands careful with the glass. <Subject 1> rocks on his heels with quick, eager breaths, eyes tracking jar to jar; he rises on his toes until his fingertips squeak against the shelf glass, taps it twice in front of a red-and-white swirl lollipop, then drops his hand and turns his shoulders toward the counter, mid-step.",
"overall_soundscape:",
"the low hum of the candy-store fridge case, a soft crinkle of cellophane wrappers, small soles squeaking on polished floorboards, muffled street traffic behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"duration_seconds": 15 "duration_seconds": 10
}, },
{ {
"id": "greenhouse_aisle", "id": "lollipop_pickup",
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Continue the exact mid-threshold motion from the previous scene without resetting it: the front wheel rolls off the concrete sill onto polished interior concrete as <Subject 1>, the same adult bicycle courier in his mustard-yellow waterproof jacket, dark teal trousers, and charcoal helmet, pushes the silver bicycle forward down the central aisle, right hand on the handlebar, left thumb still hooked under the olive messenger bag strap with the kraft paper parcel pressed against his hip. The camera keeps its stable landscape tracking glide at eye level, matching his walking pace exactly, holding his face, the parcel, and the receding fogged entrance behind him; the camera height and direction of travel never change. Photorealistic overcast rain still lays soft grey light through the glass roof, condensation trails crawling down the panes, benches of potted ferns and clay pots sliding past on both sides, aisle number cards ticking by at his shoulder. His breath stays steady with a relaxed jaw, eyes flicking to the numbers; a drop falls from the bicycle fender and bursts on the floor; the wet hems of his trousers slap lightly with each step. Approaching the wooden potting table at the aisle's end, he slows the bicycle to a crawl and his right hand leaves the handlebar, closing on the olive messenger flap.", "subject_definitions:",
"overall_soundscape: the same steady rain drumming on the glass roof continues overhead, tires humming on polished concrete, a single fender drop bursting on the floor, the distant traffic hum muffled behind the glass.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] Continuing the same shot without resetting the motion: <Subject 1> closes the last step to the counter and leans both hands on the polished wood. The camera holds a static counter-height medium shot at his eye level, the register and stacked jars soft behind his shoulder. He stretches his right hand across the counter, small fingers closing around the stick of the red-and-white swirl lollipop lying beside the till, and lifts it up in front of his eyes, the cellophane crinkling; his mouth falls open in a delighted grin and he bounces once on his toes. <Subject 2>, the heavy-set shopkeeper with short grey hair in his green apron over a rolled-sleeve white shirt, watches from behind the register and nods once, wiping his hands on a white cloth. End mid-motion: the lollipop held high in his right fist, his left hand finding the small red coin purse on its strap at his hip.",
"overall_soundscape:",
"the same low fridge-case hum and wrapper crinkle continue, the lollipop's cellophane crackling under small fingers, the till drawer rattling softly below the counter, street traffic still muffled behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"duration_seconds": 15 "duration_seconds": 10
}, },
{ {
"id": "greenhouse_delivery", "id": "pay_shopkeeper",
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward stride and the same sideways camera glide at the wooden potting table without resetting the motion, bicycle wheels rumbling to a crawl over the polished concrete. <Subject 1>, the same adult bicycle courier in his mustard-yellow waterproof jacket, dark teal trousers, and charcoal helmet, lets the silver bicycle roll its last half metre to the table, wedges the front wheel against the table leg with his left hand, unbuckles the olive messenger flap his right hand has already found, lifts the kraft paper parcel with both hands, and sets it down on the table with a soft, deliberate press. The camera glides to a stop at a calm medium-wide angle as he straightens to his full height and turns his shoulders toward the lens. He stands silently with mouth closed, then speaks. <d>[English] Package for the greenhouse, right on time.</d> He falls silently with mouth closed and lets a brief smile lift the corners of his mouth before it fades. Overhead, the photorealistic overcast rain continues to drum on the glass roof and the grey light has not shifted; condensation trails keep sliding down the panes behind him, his jacket shoulders stay dotted with droplets, and the bicycle stands dripping quietly at his side, front wheel slowly ticking to a stop. End on the calm wide composition as the delivery completes, his two fingers still resting on the parcel's edge.", "subject_definitions:",
"overall_soundscape: the same rain on glass and wet paving continues, tires creak to a stop, the kraft paper parcel thuds softly on the wooden potting table.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] Continuing the same shot without resetting the motion: <Subject 1> stands at the counter with the red-and-white swirl lollipop gripped in his right fist and the small red coin purse open in his left. The camera holds the same static counter-height medium shot, register and jars soft in the background. He tips the purse and a few coins pool into his small palm, then he reaches up on his toes to press them into <Subject 2>'s broad open hand; the heavy-set shopkeeper with short grey hair, green apron over a rolled-sleeve white shirt, closes his fingers around the coins and drops them into the wooden register drawer that springs open with a clatter, giving <Subject 1> a slow wink. <Subject 1>'s grin widens and he hops once on his toes, both fists tightening with excitement. End mid-motion: the shopkeeper's fingers still closing around the coins as the drawer slides shut, the lollipop raised high.",
"overall_soundscape:",
"the same low fridge-case hum continues, the till drawer springing open with a wooden clatter, coins clinking into the register, the shopkeeper's low chuckle, street traffic still muffled behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"duration_seconds": 15 "duration_seconds": 10
} }
] ]
} }

27
h3-plan-gate/h3_agent_workflow_examples/scene_01.json Executable file → Normal file
View File

@ -1,14 +1,27 @@
{ {
"id": "greenhouse_arrival", "id": "candy_browse",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Opening scene. <Subject 1>, the adult bicycle courier, steps off his silver bicycle at the wet gravel path outside the glass greenhouse and walks it slowly through the entrance, right hand on the handlebar, left thumb hooked under the strap of the olive messenger bag as the kraft paper parcel presses against his hip. The camera is a stable landscape tracking shot at eye level, gliding sideways at his exact walking pace, keeping his face, the parcel, and the fogged glass wall behind him in frame; the camera height and the direction of travel never change. Photorealistic overcast rainy daylight filters through the fogged panes and lays soft grey light across his mustard-yellow waterproof jacket; rain beads and long streaks run down the glass, and the wet paving glistens under the bicycle wheels. Droplets dot his charcoal helmet, his dark teal trousers darken at the shins, and each footstep squeezes a small splash between the gravel stones. He breathes steadily with a relaxed jaw, eyes flicking up to the aisle numbers as he passes benches of potted ferns and clay pots, and the kraft paper shifts faintly against the bag strap with every step. End while he is crossing the interior threshold, front wheel still rolling over the concrete sill, coat sleeve mid-swing, motion clearly still in progress.", "subject_definitions:",
"overall_soundscape: steady light rain drumming on glass roofing, wet tires crunching over wet paving, low distant traffic hum under a grey sky.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] <Subject 1> stands rooted in front of the counter-height candy shelves, both small hands hovering an inch from the glass. The camera holds a static low wide angle at his exact eye height, the ranked candy jars and bright wrapper rows sharp in shallow focus around him. <Subject 2>, the heavy-set shopkeeper with short grey hair in his green apron over a rolled-sleeve white shirt, sets a fresh jar of lollipops on the counter's far end behind the register, his big hands careful with the glass. <Subject 1> rocks on his heels with quick, eager breaths, eyes tracking jar to jar; he rises on his toes until his fingertips squeak against the shelf glass, taps it twice in front of a red-and-white swirl lollipop, then drops his hand and turns his shoulders toward the counter, mid-step.",
"overall_soundscape:",
"the low hum of the candy-store fridge case, a soft crinkle of cellophane wrappers, small soles squeaking on polished floorboards, muffled street traffic behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"_handoff": { "_handoff": {
"end_state": "The courier is crossing the interior threshold onto polished concrete, the silver bicycle's front wheel still rolling over the concrete sill, his coat sleeve mid-swing and the motion clearly still in progress, the camera gliding sideways at his exact walking pace.", "end_state": "<Subject 1> has turned from the candy shelves and is stepping toward the counter, right fingertip just off the glass, camera static at his eye height.",
"soundscape": "overall_soundscape: steady light rain drumming on glass roofing, wet tires crunching over wet paving, low distant traffic hum under a grey sky.", "soundscape": "the low hum of the candy-store fridge case, a soft crinkle of cellophane wrappers, small soles squeaking on polished floorboards, muffled street traffic behind the front glass.",
"dialogue_used": [] "dialogue_used": []
} }
} }

27
h3-plan-gate/h3_agent_workflow_examples/scene_02.json Executable file → Normal file
View File

@ -1,14 +1,27 @@
{ {
"id": "greenhouse_aisle", "id": "lollipop_pickup",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Continue the exact mid-threshold motion from the previous scene without resetting it: the front wheel rolls off the concrete sill onto polished interior concrete as <Subject 1>, the same adult bicycle courier in his mustard-yellow waterproof jacket, dark teal trousers, and charcoal helmet, pushes the silver bicycle forward down the central aisle, right hand on the handlebar, left thumb still hooked under the olive messenger bag strap with the kraft paper parcel pressed against his hip. The camera keeps its stable landscape tracking glide at eye level, matching his walking pace exactly, holding his face, the parcel, and the receding fogged entrance behind him; the camera height and direction of travel never change. Photorealistic overcast rain still lays soft grey light through the glass roof, condensation trails crawling down the panes, benches of potted ferns and clay pots sliding past on both sides, aisle number cards ticking by at his shoulder. His breath stays steady with a relaxed jaw, eyes flicking to the numbers; a drop falls from the bicycle fender and bursts on the floor; the wet hems of his trousers slap lightly with each step. Approaching the wooden potting table at the aisle's end, he slows the bicycle to a crawl and his right hand leaves the handlebar, closing on the olive messenger flap.", "subject_definitions:",
"overall_soundscape: the same steady rain drumming on the glass roof continues overhead, tires humming on polished concrete, a single fender drop bursting on the floor, the distant traffic hum muffled behind the glass.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] Continuing the same shot without resetting the motion: <Subject 1> closes the last step to the counter and leans both hands on the polished wood. The camera holds a static counter-height medium shot at his eye level, the register and stacked jars soft behind his shoulder. He stretches his right hand across the counter, small fingers closing around the stick of the red-and-white swirl lollipop lying beside the till, and lifts it up in front of his eyes, the cellophane crinkling; his mouth falls open in a delighted grin and he bounces once on his toes. <Subject 2>, the heavy-set shopkeeper with short grey hair in his green apron over a rolled-sleeve white shirt, watches from behind the register and nods once, wiping his hands on a white cloth. End mid-motion: the lollipop held high in his right fist, his left hand finding the small red coin purse on its strap at his hip.",
"overall_soundscape:",
"the same low fridge-case hum and wrapper crinkle continue, the lollipop's cellophane crackling under small fingers, the till drawer rattling softly below the counter, street traffic still muffled behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"_handoff": { "_handoff": {
"end_state": "The courier has slowed the silver bicycle to a crawl at the wooden potting table at the end of the aisle, his right hand leaving the handlebar and closing on the olive messenger flap at his hip, the kraft parcel still pressed against him, the camera still gliding at his pace.", "end_state": "<Subject 1> stands at the counter, the swirl lollipop held high in his right fist, his left hand finding the small red coin purse on its strap at his hip, camera static at counter height.",
"soundscape": "overall_soundscape: the same steady rain drumming on the glass roof continues overhead, tires humming on polished concrete, a single fender drop bursting on the floor, the distant traffic hum muffled behind the glass.", "soundscape": "the same low fridge-case hum and wrapper crinkle continue, the lollipop's cellophane crackling under small fingers, the till drawer rattling softly below the counter, street traffic still muffled behind the front glass.",
"dialogue_used": [] "dialogue_used": []
} }
} }

29
h3-plan-gate/h3_agent_workflow_examples/scene_03.json Executable file → Normal file
View File

@ -1,14 +1,27 @@
{ {
"id": "greenhouse_delivery", "id": "pay_shopkeeper",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"integrated_multimodal_description: [Shot 1] Continue the same forward stride and the same sideways camera glide at the wooden potting table without resetting the motion, bicycle wheels rumbling to a crawl over the polished concrete. <Subject 1>, the same adult bicycle courier in his mustard-yellow waterproof jacket, dark teal trousers, and charcoal helmet, lets the silver bicycle roll its last half metre to the table, wedges the front wheel against the table leg with his left hand, unbuckles the olive messenger flap his right hand has already found, lifts the kraft paper parcel with both hands, and sets it down on the table with a soft, deliberate press. The camera glides to a stop at a calm medium-wide angle as he straightens to his full height and turns his shoulders toward the lens. He stands silently with mouth closed, then speaks. <d>[English] Package for the greenhouse, right on time.</d> He falls silently with mouth closed and lets a brief smile lift the corners of his mouth before it fades. Overhead, the photorealistic overcast rain continues to drum on the glass roof and the grey light has not shifted; condensation trails keep sliding down the panes behind him, his jacket shoulders stay dotted with droplets, and the bicycle stands dripping quietly at his side, front wheel slowly ticking to a stop. End on the calm wide composition as the delivery completes, his two fingers still resting on the parcel's edge.", "subject_definitions:",
"overall_soundscape: the same rain on glass and wet paving continues, tires creak to a stop, the kraft paper parcel thuds softly on the wooden potting table.", "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair in pigtails that dresses like a baby girl",
"non_diegetic_music: N/A" "<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy build",
"summary:",
"[reference generation]",
"retention_analysis:",
"<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:",
"Warm photoreal candy-store realism, soft late-afternoon light pooling through the front windows, saturated jar-glass golds and wrapper reds under a gentle film grain.",
"Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field on the candy jars.",
"[Shot 1] Continuing the same shot without resetting the motion: <Subject 1> stands at the counter with the red-and-white swirl lollipop gripped in his right fist and the small red coin purse open in his left. The camera holds the same static counter-height medium shot, register and jars soft in the background. He tips the purse and a few coins pool into his small palm, then he reaches up on his toes to press them into <Subject 2>'s broad open hand; the heavy-set shopkeeper with short grey hair, green apron over a rolled-sleeve white shirt, closes his fingers around the coins and drops them into the wooden register drawer that springs open with a clatter, giving <Subject 1> a slow wink. <Subject 1>'s grin widens and he hops once on his toes, both fists tightening with excitement. End mid-motion: the shopkeeper's fingers still closing around the coins as the drawer slides shut, the lollipop raised high.",
"overall_soundscape:",
"the same low fridge-case hum continues, the till drawer springing open with a wooden clatter, coins clinking into the register, the shopkeeper's low chuckle, street traffic still muffled behind the front glass.",
"non_diegetic_music:",
"N/A"
], ],
"_handoff": { "_handoff": {
"end_state": "The delivery is complete: the kraft parcel rests on the wooden potting table under his two fingers, the silver bicycle stands dripping quietly at his side with its front wheel just ticked to a stop, and the camera holds the calm wide composition.", "end_state": "the shopkeeper's fingers closing around the coins as the register drawer slides shut, the lollipop raised high in <Subject 1>'s right fist, camera static at counter height.",
"soundscape": "overall_soundscape: the same rain on glass and wet paving continues, tires creak to a stop, the kraft paper parcel thuds softly on the wooden potting table.", "soundscape": "the same low fridge-case hum continues, the till drawer springing open with a wooden clatter, coins clinking into the register, the shopkeeper's low chuckle, street traffic still muffled behind the front glass.",
"dialogue_used": ["Package for the greenhouse, right on time."] "dialogue_used": []
} }
} }

View File

@ -43,10 +43,10 @@ Emit exactly this shape and no other keys:
"defaults": { "duration_seconds": 10, "steps": 20 }, "defaults": { "duration_seconds": 10, "steps": 20 },
"shots": [ "shots": [
{ "id": "scene_one_name", { "id": "scene_one_name",
"prompt": ["...the fragment's prompt lines, copied exactly..."], "prompt": ["...the fragment's six-section Ref2VA prompt lines, verbatim..."],
"duration_seconds": 10 }, "duration_seconds": 10 },
{ "id": "scene_two_name", { "id": "scene_two_name",
"prompt": ["...the fragment's prompt lines, copied exactly..."], "prompt": ["...the fragment's six-section Ref2VA prompt lines, verbatim..."],
"duration_seconds": 10 } "duration_seconds": 10 }
] ]
} }
@ -64,23 +64,22 @@ Emit exactly this shape and no other keys:
- Copy any extra "steps" or "seed" keys a fragment carries, unchanged. - Copy any extra "steps" or "seed" keys a fragment carries, unchanged.
- If any fragment is broken JSON, stop and reply exactly: BROKEN FRAGMENT <n> - If any fragment is broken JSON, stop and reply exactly: BROKEN FRAGMENT <n>
# WORKED EXAMPLE (shape only — real fragments carry full ref2va sections and # WORKED EXAMPLE (shape only — real fragments carry the full six-section prompt)
# 200-word description bodies)
Input: Input:
DEFAULTS: {"duration_seconds": 8, "steps": 20} DEFAULTS: {"duration_seconds": 8, "steps": 20}
SCENE 1 SCENE 1
{"id": "corridor_walk", "duration_seconds": 8, "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : a woman in her thirties with dark brown hair in a low bun, pale green scrubs, white sneakers.", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Photorealistic cool fluorescent lighting, soft clinical palette.", "[Shot 1] <Subject 1>, the same night nurse from her reference image in her pale green scrub top, matching scrub trousers, and white sneakers, her nude hands gripping the trolley handle, pushes the trolley down the corridor.", "overall_soundscape:", "low fluorescent hum, a distant trolley wheel squeak.", "non_diegetic_music:", "N/A"], "_handoff": {"end_state": "She is mid-stride at the far end.", "soundscape": "low fluorescent hum, a distant trolley wheel squeak.", "dialogue_used": []}} {"id": "corridor_walk", "duration_seconds": 8, "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : the night nurse from the reference photo - pale green scrubs, white sneakers, brown hair in a low clip", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Cool fluorescent corridor realism.", "[Shot 1] <Subject 1>, the same night nurse in her pale green scrubs and white sneakers, pushes the trolley down the corridor.", "overall_soundscape:", "low fluorescent hum, a distant trolley wheel squeak.", "non_diegetic_music:", "N/A"], "_handoff": {"end_state": "She is mid-stride at the far end.", "soundscape": "low fluorescent hum, a distant trolley wheel squeak.", "dialogue_used": []}}
SCENE 2 SCENE 2
{"id": "door_stop", "duration_seconds": 8, "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : a woman in her thirties with dark brown hair in a low bun, pale green scrubs, white sneakers.", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Photorealistic cool fluorescent lighting, soft clinical palette.", "[Shot 1] Continue the same walk as <Subject 1>, the same night nurse from her reference image in her pale green scrub top, matching scrub trousers, and white sneakers, her nude hands opening off the trolley handle, and stops the trolley at the blue door.", "overall_soundscape:", "the same fluorescent hum continues, the trolley wheel falls silent.", "non_diegetic_music:", "N/A"], "_handoff": {"end_state": "Her hand rests on the push bar.", "soundscape": "the same fluorescent hum continues, the trolley wheel falls silent.", "dialogue_used": []}} {"id": "door_stop", "duration_seconds": 8, "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : the night nurse from the reference photo - pale green scrubs, white sneakers, brown hair in a low clip", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Cool fluorescent corridor realism.", "[Shot 1] Continue the same walk as <Subject 1>, the same night nurse in her pale green scrubs, stops the trolley at the blue door.", "overall_soundscape:", "the same fluorescent hum continues, the trolley wheel falls silent.", "non_diegetic_music:", "N/A"], "_handoff": {"end_state": "Her hand rests on the push bar.", "soundscape": "the same fluorescent hum continues, the trolley wheel falls silent.", "dialogue_used": []}}
Correct reply: Correct reply:
{"prompt_prefix": [""], "defaults": {"duration_seconds": 8, "steps": 20}, "shots": [ {"prompt_prefix": [""], "defaults": {"duration_seconds": 8, "steps": 20}, "shots": [
{"id": "corridor_walk", "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : a woman in her thirties with dark brown hair in a low bun, pale green scrubs, white sneakers.", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Photorealistic cool fluorescent lighting, soft clinical palette.", "[Shot 1] <Subject 1>, the same night nurse from her reference image in her pale green scrub top, matching scrub trousers, and white sneakers, her nude hands gripping the trolley handle, pushes the trolley down the corridor.", "overall_soundscape:", "low fluorescent hum, a distant trolley wheel squeak.", "non_diegetic_music:", "N/A"], "duration_seconds": 8}, {"id": "corridor_walk", "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : the night nurse from the reference photo - pale green scrubs, white sneakers, brown hair in a low clip", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Cool fluorescent corridor realism.", "[Shot 1] <Subject 1>, the same night nurse in her pale green scrubs and white sneakers, pushes the trolley down the corridor.", "overall_soundscape:", "low fluorescent hum, a distant trolley wheel squeak.", "non_diegetic_music:", "N/A"], "duration_seconds": 8},
{"id": "door_stop", "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : a woman in her thirties with dark brown hair in a low bun, pale green scrubs, white sneakers.", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Photorealistic cool fluorescent lighting, soft clinical palette.", "[Shot 1] Continue the same walk as <Subject 1>, the same night nurse from her reference image in her pale green scrub top, matching scrub trousers, and white sneakers, her nude hands opening off the trolley handle, and stops the trolley at the blue door.", "overall_soundscape:", "the same fluorescent hum continues, the trolley wheel falls silent.", "non_diegetic_music:", "N/A"], "duration_seconds": 8} {"id": "door_stop", "prompt": ["subject_definitions:", "<Subject 1> binds to <Image 1> : the night nurse from the reference photo - pale green scrubs, white sneakers, brown hair in a low clip", "summary:", "[reference generation]", "retention_analysis:", "<Subject 1>: fully_preserved", "detailed_description:", "Cool fluorescent corridor realism.", "[Shot 1] Continue the same walk as <Subject 1>, the same night nurse in her pale green scrubs, stops the trolley at the blue door.", "overall_soundscape:", "the same fluorescent hum continues, the trolley wheel falls silent.", "non_diegetic_music:", "N/A"], "duration_seconds": 8}
]} ]}
Note what the example shows: the "_handoff" objects vanished, every prompt string is Note what the example shows: the "_handoff" objects vanished, every prompt string is
@ -106,9 +105,9 @@ DEFAULTS block.
`plan_json_input` (a non-empty connected value overrides the visual editor's plan for `plan_json_input` (a non-empty connected value overrides the visual editor's plan for
that execution and receives the identical validation). that execution and receives the identical validation).
- **QA after assembly:** `python scripts/verify_plan_rules.py plan plan.json` re-runs the - **QA after assembly:** `python scripts/verify_plan_rules.py plan plan.json` re-runs the
node-rule replica plus the H3 ref2va six-section (or legacy t2va three-line, detected node-rule replica plus the six-section Ref2VA checks (section order, definition forms,
per shot) / word-budget / handoff-free checks on the final file. The deterministic refmod isolation, retention coverage) on the
merger runs all of this automatically. final file. The deterministic merger runs all of this automatically.
- **Token footprint (measured, o200k proxy):** the prompt is ~1.2k tokens; input is - **Token footprint (measured, o200k proxy):** the prompt is ~1.2k tokens; input is
roughly the sum of all fragments (~0.5k per scene); output is nearly the same size roughly the sum of all fragments (~0.5k per scene); output is nearly the same size
again. A 12-scene assembly turn is ~12k tokens — fine at 128k ctx; if you ever run again. A 12-scene assembly turn is ~12k tokens — fine at 128k ctx; if you ever run

View File

@ -51,22 +51,22 @@ Emit exactly this shape and no other keys:
{ "id": "scene_one", { "id": "scene_one",
"prompt": [ "prompt": [
"subject_definitions:", "subject_definitions:",
"<Subject 1> ...one of the three definition forms (REF2VA PROMPT 1)...", "<Subject 1> is a ...",
"summary:", "summary:",
"[reference generation]", "[reference generation]",
"retention_analysis:", "retention_analysis:",
"<Subject 1>: fully_preserved", "<Subject 1>: fully_preserved",
"detailed_description:", "detailed_description:",
"<1-2 style sentences from the story's look>", "<1-2 style sentences>",
"[Shot 1] ...the full scene composition...", "[Shot 1] ...",
"overall_soundscape:", "overall_soundscape:",
"...the sound the scene itself makes...", "<the scene's own sound>",
"non_diegetic_music:", "non_diegetic_music:",
"N/A" "N/A"
], ],
"duration_seconds": 10 }, "duration_seconds": 10 },
{ "id": "scene_two", { "id": "scene_two",
"prompt": [ "...the same ref2va sections with this scene's content..." ], "prompt": [ "...the same six AUDIO sections with this scene's content..." ],
"duration_seconds": 10 } "duration_seconds": 10 }
] ]
} }
@ -74,8 +74,7 @@ Emit exactly this shape and no other keys:
# TOP-LEVEL FIELDS # TOP-LEVEL FIELDS
- "prompt_prefix": always exactly [""] — one empty string, nothing else. Subject identity, - "prompt_prefix": always exactly [""] — one empty string, nothing else. Subject identity,
wardrobe, art style, camera style, and continuity rules belong inside every scene's wardrobe, art style, camera style, and continuity rules belong inside every scene's
subject_definitions and detailed_description sections instead, so each scene is detailed_description body instead, so each scene is self-sufficient.
self-sufficient.
- "defaults": object with exactly two keys: "duration_seconds" (positive number) and "steps" - "defaults": object with exactly two keys: "duration_seconds" (positive number) and "steps"
(integer). Use 10 and 20 unless the user says otherwise. (integer). Use 10 and 20 unless the user says otherwise.
- "shots": array of shot objects in story order. Usually 3 to 12 shots. If the user names a - "shots": array of shot objects in story order. Usually 3 to 12 shots. If the user names a
@ -85,8 +84,8 @@ Emit exactly this shape and no other keys:
Each shot object has exactly these keys: Each shot object has exactly these keys:
- "id": unique short snake_case name. Lowercase letters, digits, underscores only. - "id": unique short snake_case name. Lowercase letters, digits, underscores only.
Every id must be different. Example: "kitchen_argument". Every id must be different. Example: "kitchen_argument".
- "prompt": array of strings in the REF2VA PROMPT order given below — one line per - "prompt": array of strings compiling to the six-section AUDIO prompt given below. Never
element, section headers on their own line. Never empty. empty. The label lines stand alone; content lines follow them.
- "duration_seconds": positive number between 2 and 15. Omit it to inherit the default. - "duration_seconds": positive number between 2 and 15. Omit it to inherit the default.
Only when the user explicitly asks, a shot may also include: Only when the user explicitly asks, a shot may also include:
- "steps": integer from 1 to 10000. - "steps": integer from 1 to 10000.
@ -124,13 +123,11 @@ Shots are flat. A shot never contains a nested object.
# PROMPT CRAFT # PROMPT CRAFT
- "prompt_prefix" stays empty, exactly [""]. Never put identity, wardrobe, style, or - "prompt_prefix" stays empty, exactly [""]. Never put identity, wardrobe, style, or
continuity text in it. Every scene's prompt must be self-sufficient: continuity text in it. Every scene's detailed_description body must be
subject_definitions anchors the <Subject 1> tag's identity in one of its three self-sufficient: introduce or re-state the <Subject 1> tag, wardrobe, look, and visual
forms and the [Shot 1] body carries the full look woven into the action in every style in every scene, because nothing is shared from a prefix. A user-refmod subject is
scene, because nothing is shared from a prefix or between scenes (a refmod the exception — bare "<Subject N>" token outside subject_definitions (AUDIO below).
subject excepted: its verbatim definition carries the identity and look, and the - The detailed_description body of every scene must:
body references only `<Subject N>`).
- The [Shot 1] body of every scene's detailed_description must:
1. Continue the exact action, pose, camera direction, and lighting from the previous 1. Continue the exact action, pose, camera direction, and lighting from the previous
scene's [Shot 1] body. Scene 1 instead sets the opening. scene's [Shot 1] body. Scene 1 instead sets the opening.
2. Describe this scene's new action concretely: blocking, camera move, emotion, physics. 2. Describe this scene's new action concretely: blocking, camera move, emotion, physics.
@ -139,18 +136,11 @@ Shots are flat. A shot never contains a nested object.
- Compose every scene from the same eight elements, and never chase a word count — - Compose every scene from the same eight elements, and never chase a word count —
chase the elements. A description that genuinely carries them comes out long enough chase the elements. A description that genuinely carries them comes out long enough
on its own; padding to a number is as much a failure as a thin one-liner: on its own; padding to a number is as much a failure as a thin one-liner:
1. SUBJECT ANCHOR — subject_definitions anchors identity only, one line per subject 1. SUBJECT ANCHOR — <Subject 1> restated. For no-image and binds subjects: identity,
in exactly one of three forms (REF2VA PROMPT below): physical traits only when every aspect of how they look, wardrobe head to toe, hair, build — naming every body
the character has no reference image ("is a ... with ..." — age, hair, eyes, part the clothing leaves uncovered, using "nude" for uncovered skin (PROMPT CRAFT).
build), "binds to <Image N> : what the image actually shows" when one is wired For a user-refmod subject: the bare "<Subject 1>" token only (REFMOD ISOLATION) —
(you are multimodal — look at it), or the user's refmod carried through verbatim the refmod already pins the look, never re-describe it.
("identity: ...; voice: ..."). For a no-image or binds subject the full look
lives in the [Shot 1] body, woven into the action as the scene exposes it:
hair, build, and face as they move, every garment where the action touches it,
footwear where steps land, and every body part the clothing leaves uncovered
named "nude" (PROMPT CRAFT) where the scene shows or uses that skin. A refmod
subject's body references only its `<Subject N>` label — the refmod text never
leaves subject_definitions.
2. STYLE ANCHOR — the one consistent look: medium, lighting quality, palette. 2. STYLE ANCHOR — the one consistent look: medium, lighting quality, palette.
3. CONTINUITY ANCHOR — open inside the previous scene's end state (scene 1 sets the 3. CONTINUITY ANCHOR — open inside the previous scene's end state (scene 1 sets the
opening) and name what carries over: the stride, the camera direction, the light, opening) and name what carries over: the stride, the camera direction, the light,
@ -213,22 +203,16 @@ Shots are flat. A shot never contains a nested object.
frame"; not "the lights dim" but "the overhead strip flickers once, steadies, then frame"; not "the lights dim" but "the overhead strip flickers once, steadies, then
fades until only the doorway lamp lights the room". A summarized change or movement fades until only the doorway lamp lights the room". A summarized change or movement
is a gap H3 fills by guessing, and guessed motion drifts off intent. is a gap H3 fills by guessing, and guessed motion drifts off intent.
- Describe the complete look in every scene, inside the [Shot 1] body — subject - Describe the complete look in every scene for no-image and binds subjects: every aspect
definitions never carry the scene's look; they only anchor identity in one of of how the person looks — face, hair, build, each garment head to toe, footwear —
their three forms (a refmod subject is the exception: its definition carries the because nothing carries over between scenes. Any body part the clothing leaves uncovered
look, so its body references only `<Subject N>`). Weave the full must be named and called "nude", never "bare": not "a man wearing a white tank-top" but
head-to-toe appearance into the action according to what is going on in the scene: "a man wearing a white tank-top with nude arms"; likewise nude legs under shorts, nude
face, hair, and build as the body moves, each garment where the action touches it shoulders under a sleeveless top, nude feet in sandals. "Nude" is the reliable
(a cuff sliding back, a hem swaying, fabric darkening), footwear where steps land. skin-state word for uncovered body parts; "bare" is not. A user-refmod subject never
Any body part the clothing leaves uncovered is named and called "nude", never gets this treatment — the bare "<Subject N>" token only (REFMOD ISOLATION).
"bare", stated where the scene exposes that skin: not "a man wearing a white - One string per line in the prompt array. A shot's array holds exactly the six AUDIO
tank-top" but "a man wearing a white tank-top with nude arms"; likewise nude legs sections below; never add blank entries, extra prose lines, or a second section label.
under shorts, nude shoulders under a sleeveless top, nude feet in sandals. When the
action changes what is worn or shown — a jacket coming off, a leg crossing over —
state it as part of that action, not as a static inventory. "Nude" is the reliable
skin-state word for uncovered body parts; "bare" is not.
- One line per element in the prompt array. A shot's array holds exactly the six ref2va
sections below; never add blank entries, extra prose lines, or a second section header.
- Useful continuity phrases: "Continue the same stride", "without resetting the motion", - Useful continuity phrases: "Continue the same stride", "without resetting the motion",
"keep the incoming camera direction", "same lighting", "no cut". "keep the incoming camera direction", "same lighting", "no cut".
- <Subject 1> (and <Subject 2> when the story casts a second recurring person) are - <Subject 1> (and <Subject 2> when the story casts a second recurring person) are
@ -238,49 +222,35 @@ Shots are flat. A shot never contains a nested object.
references. references.
- Keep one consistent style across the whole plan. Never switch style mid-plan. - Keep one consistent style across the whole plan. Never switch style mid-plan.
# REF2VA PROMPT (every shot prompt is the six native H3 reference sections) # AUDIO (every shot prompt is the six-section Ref2VA prompt)
- Each scene is rendered as ONE H3 generation built from the subject reference images - Each scene is rendered as ONE H3 generation with reference conditioning, and the H3 model
the user has wired, and the H3 model reserves special tokens for six exact section reads six exact section labels. Give every shot a prompt array with exactly these six
labels. Give every shot a prompt array with exactly these sections, in this order, sections, in this order (one string per line; the label lines stand alone):
one line per element — the header on its own line, the content on the following 1. "subject_definitions:" — followed by one line per subject. Each line is exactly one of
element: the three pinning forms: "<Subject N> is a <age> ... with <hair>, <eyes> and <build>"
1. "subject_definitions:" then one line per subject — each line is exactly one of (no reference image — physical traits only: never clothing, never a nude/bare skin
three forms: state), "<Subject N> binds to <Image N> : <what the image shows>" (reference image —
a. No reference image for this character: physical traits only — "<Subject 1> faithful to it), or "<Subject N> identity: <user text>; voice: <user text>" (user
is a 20 year old woman with blonde hair, blue eyes and a medium build." Age, refmod — the user's tested wording verbatim, the story's &...& wrapper stripped;
hair, eyes, build. Never clothing, never a nude/bare skin state — the look is exempt from the no-image bans). No form carries a length cap. REFMOD ISOLATION starts
woven into the [Shot 1] body instead. here: a refmod subject's wording appears nowhere else in the prompt — the rest of the
b. A reference image is wired for this character: "<Subject 1> binds to <Image 1> shot references the subject by the bare "<Subject N>" token only, never re-describing
: <what the image actually shows>." You are multimodal: look at <Image 1> and the pinned look or voice, because the workflow injects the reference blocks itself.
describe what you see — age, hair, eyes, build, what they wear in the image. 2. "summary:" — followed by exactly "[reference generation]" and nothing else. Every
Never invent what you cannot see. chain shot is its own reference generation.
c. The user wrote a refmod — the story carries a `characters:` block with the 3. "retention_analysis:" — followed by one "<Subject N>: fully_preserved" line per
name as `sam:&identity: <identity text>; voice: <voice text>&`. Strip the & defined subject, so every defined identity is kept across the chain.
marks and carry the text through VERBATIM: "<Subject 1> identity: <their 4. "detailed_description:" — followed by 1-2 style sentences (the one consistent look:
identity text>; voice: <their voice text>". Never rewrite, trim, reorder, or medium, lighting, palette, camera feel), then the "[Shot 1]" body: subject anchor,
"improve" a refmod line — it is the user's tested wording and overrides every continuity anchor, blocking, camera work, performance, physical texture, exit state,
other rule about definitions (the no-image clothing and skin-state bans and any <d> dialogue. The bracketed number is always [Shot 1] — every scene is its
never apply to a refmod line). own single-shot call.
Define <Subject 2> the same way when the story casts a second recurring person. 5. "overall_soundscape:" — the sound the scene itself makes: weather, room tone,
Never define a label twice; never define a label you do not use. footsteps, surfaces, machines, animals, crowd, and diegetic music always named with its
2. "summary:" then "[reference generation]" — exactly this directive on every scene, source inside the scene, e.g. "a battery radio on the shelf plays soft instrumental
nothing else in the section. jazz".
3. "retention_analysis:" then one line per defined label: "<Subject 1>: 6. "non_diegetic_music:" — write N/A unless the user asks for background music that
fully_preserved" (alternatives only when the scene really calls for them: nobody in the scene can hear; only then give one short clause describing it.
partially_preserved, attribute_transfer, weak_reference; default
fully_preserved).
4. "detailed_description:" then 1-2 style sentences (medium, lighting quality,
palette — the look the whole plan shares), then the "[Shot 1] ..." body — one
line holding the subject references, visual style cues, all action, camera
work, continuity cues, and any <d> dialogue. Every scene is its own single-shot
call, so the bracketed number is always [Shot 1].
5. "overall_soundscape:" then the sound the scene itself makes: weather, room tone,
footsteps, surfaces, machines, animals, crowd, and diegetic music always named
with its source inside the scene, e.g. "a battery radio on the shelf plays soft
instrumental jazz".
6. "non_diegetic_music:" then "N/A" — write N/A unless the user asks for background
music that nobody in the scene can hear; only then give one short clause
describing it.
- Scene 1 establishes the soundscape. Later scenes keep the same audio bed and change only - Scene 1 establishes the soundscape. Later scenes keep the same audio bed and change only
what the picture changes; say "the same rain continues" instead of inventing new weather. what the picture changes; say "the same rain continues" instead of inventing new weather.
- The soundscape must match what the picture shows. No visible or implied source, no sound. - The soundscape must match what the picture shows. No visible or implied source, no sound.
@ -305,9 +275,9 @@ Shots are flat. A shot never contains a nested object.
- Approved markers: [English] [French] [Spanish] [German] [Italian] [Portuguese] - Approved markers: [English] [French] [Spanish] [German] [Italian] [Portuguese]
[Chinese] [Japanese] [Korean] [Arabic]. Another real language name in brackets also [Chinese] [Japanese] [Korean] [Arabic]. Another real language name in brackets also
works, like [Hindi]. Never use [unclear]. works, like [Hindi]. Never use [unclear].
- Keep each <d>...</d> span inside ONE prompt array entry, always the [Shot 1] element - Keep each <d>...</d> span inside ONE prompt array entry, always the detailed_description
of detailed_description. Never split a span across two entries, never put dialogue in [Shot 1] body. Never split a span across two entries, never put
the soundscape section, and never put a raw line break inside a span. dialogue in a soundscape line, and never put a raw line break inside a span.
- Bracket every dialogue span with mouth-state cues, or H3 chatters gibberish around - Bracket every dialogue span with mouth-state cues, or H3 chatters gibberish around
the line. Use the phrase "silently with mouth" plus a concrete position: the line. Use the phrase "silently with mouth" plus a concrete position:
1. Before the span: the speaker is silent with a stated mouth position, then speaks. 1. Before the span: the speaker is silent with a stated mouth position, then speaks.
@ -333,8 +303,8 @@ Correct reply:
"id": "greenhouse_arrival", "id": "greenhouse_arrival",
"prompt": [ "prompt": [
"subject_definitions:", "subject_definitions:",
"<Subject 1> binds to <Image 1> : a lean adult bicycle courier with short "<Subject 1> is an adult man with a weathered face, short dark hair and a
dark brown hair and a mustard-yellow waterproof jacket.", solid working build",
"summary:", "summary:",
"[reference generation]", "[reference generation]",
"retention_analysis:", "retention_analysis:",
@ -342,30 +312,27 @@ Correct reply:
"detailed_description:", "detailed_description:",
"Photorealistic overcast rainy daylight, soft grey palette, stable landscape "Photorealistic overcast rainy daylight, soft grey palette, stable landscape
tracking shots at eye level.", tracking shots at eye level.",
"[Shot 1] Opening scene. <Subject 1>, the lean-built adult bicycle courier "[Shot 1] Opening scene. <Subject 1>, the
with short dark brown hair, steps off his silver bicycle at the wet gravel adult bicycle courier, steps off his silver bicycle at the wet gravel path
path outside the glass greenhouse and walks it slowly through the entrance, outside the glass greenhouse and walks it slowly through the entrance, right
right hand on the handlebar, the nude hand below his jacket cuff wrapped hand on the handlebar, left thumb hooked under the strap of the olive
around the grip, left thumb hooked under the strap of the olive messenger messenger bag as the kraft paper parcel presses against his hip. The camera
bag as the kraft paper parcel presses against his hip. The camera is a is a stable landscape tracking shot at eye level, gliding sideways at his
stable landscape tracking shot at eye level, gliding sideways at his exact exact walking pace, keeping his face, the parcel, and the fogged glass wall
walking pace, keeping his face, the parcel, and the fogged glass wall
behind him in frame; the camera height and the direction of travel never behind him in frame; the camera height and the direction of travel never
change. Photorealistic overcast rainy daylight filters through the fogged change. Photorealistic overcast rainy daylight filters through the fogged panes and lays
panes and lays soft grey light across his mustard-yellow waterproof jacket soft grey light across his mustard-yellow waterproof jacket; rain beads and
zipped to the collar; rain beads and long streaks run down the glass, and long streaks run down the glass, and the wet paving glistens under the
the wet paving glistens under the bicycle wheels. Droplets dot his charcoal bicycle wheels. Droplets dot his charcoal helmet, his dark teal trousers
helmet, its straps holding his nude ears close, his dark teal trousers darken at the shins, and each footstep squeezes a small splash between the
darken at the shins, and his silver-grey sneakers squeeze small splashes gravel stones. He breathes steadily with a relaxed jaw, eyes flicking up to
from the gravel with each step. He breathes steadily with a relaxed jaw, the aisle numbers as he passes benches of potted ferns and clay pots, and
eyes flicking up to the aisle numbers as he passes benches of potted ferns the kraft paper shifts faintly against the bag strap with every step. End
and clay pots, and the kraft paper shifts faintly against the bag strap while he is crossing the interior threshold, front wheel still rolling over
with every step. End while he is crossing the interior threshold, front the concrete sill, coat sleeve mid-swing, motion clearly still in progress.",
wheel still rolling over the concrete sill, coat sleeve mid-swing, motion
clearly still in progress.",
"overall_soundscape:", "overall_soundscape:",
"steady light rain drumming on glass roofing, wet tires crunching over wet "steady light rain drumming on glass roofing, wet tires
paving, low distant traffic hum under a grey sky.", crunching over wet paving, low distant traffic hum under a grey sky.",
"non_diegetic_music:", "non_diegetic_music:",
"N/A" "N/A"
], ],
@ -375,8 +342,8 @@ Correct reply:
"id": "greenhouse_delivery", "id": "greenhouse_delivery",
"prompt": [ "prompt": [
"subject_definitions:", "subject_definitions:",
"<Subject 1> binds to <Image 1> : a lean adult bicycle courier with short "<Subject 1> is an adult man with a weathered face, short dark hair and a
dark brown hair and a mustard-yellow waterproof jacket.", solid working build",
"summary:", "summary:",
"[reference generation]", "[reference generation]",
"retention_analysis:", "retention_analysis:",
@ -384,30 +351,30 @@ Correct reply:
"detailed_description:", "detailed_description:",
"Photorealistic overcast rainy daylight, soft grey palette, stable landscape "Photorealistic overcast rainy daylight, soft grey palette, stable landscape
tracking shots at eye level.", tracking shots at eye level.",
"[Shot 1] Continue the same forward stride and the same sideways camera move "[Shot 1] Continue the same forward stride
through the already-crossed threshold without resetting the motion, wet and the same sideways camera move through the already-crossed threshold
gravel giving way to polished concrete underfoot. <Subject 1>, the same without resetting the motion, wet gravel giving way to polished concrete
lean-built courier with his short dark brown hair under the charcoal underfoot. <Subject 1>, the same adult bicycle courier in his mustard-yellow
helmet, in his mustard-yellow waterproof jacket, dark teal trousers, and waterproof jacket, dark teal trousers, and charcoal helmet, rolls the silver
silver-grey sneakers, rolls the silver bicycle to the wooden potting table bicycle to the wooden potting table at the end of the aisle, wedges the front
at the end of the aisle, wedges the front wheel against the table leg with wheel against the table leg with his left hand, unbuckles the olive messenger
his left hand, unbuckles the olive messenger flap with the nude hands below flap, lifts the kraft paper parcel with both
his jacket cuffs, lifts the kraft paper parcel with both hands, and sets it hands, and sets it down on the table with a soft, deliberate press. The
down on the table with a soft, deliberate press. camera glides to a stop at a calm medium-wide angle as he straightens to his
The camera glides to a stop at a calm medium-wide angle as he straightens to full height and turns his shoulders toward the lens. He stands silently with
his full height and turns his shoulders toward the lens. He stands silently mouth closed, then speaks. <d>[English] Package for the greenhouse, right on
with mouth closed, then speaks. <d>[English] Package for the greenhouse, time.</d> He falls silently with mouth closed and lets a brief smile lift the
right on time.</d> He falls silently with mouth closed and lets a brief smile corners of his mouth before it fades. Overhead, the photorealistic overcast rain
lift the corners of his mouth before it fades. Overhead, the photorealistic continues to drum on the glass roof and the grey light has not shifted;
overcast rain continues to drum on the glass roof and the grey light has not condensation trails keep
shifted; condensation trails keep sliding down the panes behind him, his sliding down the panes behind him, his jacket shoulders stay dotted with
jacket shoulders stay dotted with droplets, and the bicycle stands dripping droplets, and the bicycle stands dripping quietly at his side, front wheel
quietly at his side, front wheel slowly ticking to a stop. End on the calm slowly ticking to a stop. End on the calm wide composition as the delivery
wide composition as the delivery completes, his two fingers still resting on completes, his two fingers still resting on the parcel's edge.",
the parcel's edge.",
"overall_soundscape:", "overall_soundscape:",
"the same rain on glass and wet paving continues, tires creak to a stop, the "the same rain on glass and wet paving continues, tires
kraft paper parcel thuds softly on the wooden potting table.", creak to a stop, the kraft paper parcel thuds softly on the wooden potting
table.",
"non_diegetic_music:", "non_diegetic_music:",
"N/A" "N/A"
], ],
@ -444,17 +411,20 @@ Correct reply:
wardrobe — is spelled out step by step the same way, never summarized. No wardrobe — is spelled out step by step the same way, never summarized. No
common action is only mentioned — unbuttoning, diaper changes, lamp switches, common action is only mentioned — unbuttoning, diaper changes, lamp switches,
doors, straps, and hand-offs all appear as their on-screen steps. doors, straps, and hand-offs all appear as their on-screen steps.
9. Every shot's prompt array carries the six ref2va sections in canonical order — 9. Every shot's prompt array compiles to the six AUDIO sections in this order:
subject_definitions / summary / retention_analysis / detailed_description / subject_definitions (one pinning-form line per subject) / summary (exactly
overall_soundscape / non_diegetic_music — each header on its own line, and "[reference generation]") / retention_analysis (one line per defined subject) /
summary is exactly "[reference generation]". detailed_description (1-2 style sentences, then the "[Shot 1]" body) /
overall_soundscape / non_diegetic_music. Refmod isolation holds: outside
subject_definitions a refmod subject is only the bare "<Subject N>" token.
10. overall_soundscape matches what the picture shows; the same audio bed continues across 10. overall_soundscape matches what the picture shows; the same audio bed continues across
scenes; non_diegetic_music: is N/A unless the user asked for a score; no lyrics anywhere scenes; non_diegetic_music: is N/A unless the user asked for a score; no lyrics anywhere
unless the user wrote them. unless the user wrote them.
11. There are no <d> spans at all unless the user's story provided spoken words; if it 11. There are no <d> spans at all unless the user's story provided spoken words; if it
did, the text inside every <d>...</d> matches the user's words exactly. did, the text inside every <d>...</d> matches the user's words exactly.
12. Every spoken line is wrapped as <d>[Language] words</d> inside the [Shot 1] body, 12. Every spoken line is wrapped as <d>[Language] words</d> inside the
with a "silently with mouth" cue before and after the span. detailed_description [Shot 1] body, with a "silently with mouth" cue before and
after the span.
``` ```
--- ---
@ -477,7 +447,7 @@ A retired boxer teaches his granddaughter to dance in their cluttered kitchen.
| LLM settings | Temperature 0.3–0.7. If your server supports `response_format: {"type": "json_object"}` or a grammar/JSON-mode constraint, enable it — this prompt is fully compatible with it. | | LLM settings | Temperature 0.3–0.7. If your server supports `response_format: {"type": "json_object"}` or a grammar/JSON-mode constraint, enable it — this prompt is fully compatible with it. |
| LLM context window | The prompt itself is ~3.8k tokens (measured); with the 200-word-per-scene description minimum, generated plans add ~0.5k tokens per scene — a 12-shot plan turn is ~9.5k tokens. Set ctx ≥ 16384 for 12-shot plans; 8192 still covers ≤8 shots. (User runs at 128k — far more than enough for any plan size, including 128-shot maximums.) | | LLM context window | The prompt itself is ~3.8k tokens (measured); with the 200-word-per-scene description minimum, generated plans add ~0.5k tokens per scene — a 12-shot plan turn is ~9.5k tokens. Set ctx ≥ 16384 for 12-shot plans; 8192 still covers ≤8 shots. (User runs at 128k — far more than enough for any plan size, including 128-shot maximums.) |
| Dialogue format | Spoken lines go in `<d>[Language] words</d>` (case-sensitive tags, marker + one space; language list verified against the node's own `h3_prompt_schema_core.mjs` and `h3_prompt_completion_core.mjs`). Each span is bracketed by "silently with mouth {position}" cues before and after — the repo's `AUDIO_AND_CONTINUITY.md` and MiniMax's official H3 guide both require describing mouth state in the prompt, and without it H3 chirps/gibberishes around dialogue. | | Dialogue format | Spoken lines go in `<d>[Language] words</d>` (case-sensitive tags, marker + one space; language list verified against the node's own `h3_prompt_schema_core.mjs` and `h3_prompt_completion_core.mjs`). Each span is bracketed by "silently with mouth {position}" cues before and after — the repo's `AUDIO_AND_CONTINUITY.md` and MiniMax's official H3 guide both require describing mouth state in the prompt, and without it H3 chirps/gibberishes around dialogue. |
| Soundscape / H3 sections | Every shot prompt is written in the **native H3 ref2va (full-reference) template**: `subject_definitions:` → `summary:` (`[reference generation]`) → `retention_analysis:` → `detailed_description:` (1–2 style sentences, then `[Shot 1] …`) → `overall_soundscape:` → `non_diegetic_music:` — order fixed, verified against `web/h3_prompt_schema_core.mjs` (`H3_REFERENCE_SECTIONS`, kept aligned with H3 tokenizer special tokens per ComfyUI PR #15808). The chain node does **not** add these labels — it compiles each shot as `prompt_prefix` + blank line + scene prompt verbatim (`chain_nodes.py`), so the labels must be inside the JSON prompt lines, headers on their own line. `<Subject N>` definitions resolve against the subject reference images the user wires (Scheduled Ref2VA / `native_ref2va`). Diegetic music (in-scene source) goes in the soundscape section; score goes in `non_diegetic_music` only when the user asks. Audio is heard only if the node's Generation Profile actually generates it (e.g. "Generate audio"). | | Soundscape / H3 sections | Every shot prompt is written in the **native H3 Ref2VA template**: `subject_definitions:` → `summary: [reference generation]` → `retention_analysis:` → `detailed_description:` → `overall_soundscape:` → `non_diegetic_music:` — order fixed, verified against `web/h3_prompt_schema_core.mjs` (`H3_REFERENCE_SECTIONS`, kept aligned with H3 tokenizer special tokens per ComfyUI PR #15808). The chain node does **not** add these labels — it compiles each shot as `prompt_prefix` + blank line + scene prompt verbatim (`chain_nodes.py`), so the labels must be inside the JSON prompt lines. Diegetic music (in-scene source) goes in the soundscape line; score goes in `non_diegetic_music` only when the user asks. Audio is heard only if the node's Generation Profile actually generates it (e.g. "Generate audio"). |
| Resolution / aspect | **Not a plan-JSON field.** `_normalize_plan` takes width/height only from the Plan node's `width`/`height` widgets (defaults 960×544, step 32, max 4096); a JSON `compatibility` block is ignored and per-shot `resolution` applies only via chapters (which this prompt forbids). To change aspect: set the Plan widgets (e.g. 960×544 ≈16:9, 896×672 4:3, 544×960 portrait) **and** connect the Plan/Current Shot width & height outputs to the H3 conditioning node and sampler latent — README: "Connect Current Shot width/height to the H3 conditioning node." Disconnected widgets leave whatever value (often square) is baked into the saved workflow. | | Resolution / aspect | **Not a plan-JSON field.** `_normalize_plan` takes width/height only from the Plan node's `width`/`height` widgets (defaults 960×544, step 32, max 4096); a JSON `compatibility` block is ignored and per-shot `resolution` applies only via chapters (which this prompt forbids). To change aspect: set the Plan widgets (e.g. 960×544 ≈16:9, 896×672 4:3, 544×960 portrait) **and** connect the Plan/Current Shot width & height outputs to the H3 conditioning node and sampler latent — README: "Connect Current Shot width/height to the H3 conditioning node." Disconnected widgets leave whatever value (often square) is baked into the saved workflow. |
| Caching behavior | The plan omits `prompt_seed_mode` and `chapters`, so the node's `IS_CHANGED` stays cache-stable between identical queue runs. | | Caching behavior | The plan omits `prompt_seed_mode` and `chapters`, so the node's `IS_CHANGED` stays cache-stable between identical queue runs. |
@ -491,29 +461,22 @@ A retired boxer teaches his granddaughter to dance in their cluttered kitchen.
- **Exact frame control:** if you ever need frame-exact scenes, allow `"length"` with values - **Exact frame control:** if you ever need frame-exact scenes, allow `"length"` with values
on the 17k+5 grid (5, 22, 39, 56 … 3592) — but for a local model, `duration_seconds` is far on the 17k+5 grid (5, 22, 39, 56 … 3592) — but for a local model, `duration_seconds` is far
safer; the node rounds up for you. safer; the node rounds up for you.
- **Reference-tag prompts:** `<Subject N>` definitions are the ref2va reference format - **Reference-tag prompts:** the tag rule (`<Picture N>` / `<Video N>` / `<Audio N>` only when
and resolve against the subject reference images wired in ComfyUI (Scheduled the user describes references) matches how Scheduled Ref2VA aliases compile; tags that don't
Ref2VA / `native_ref2va`). A character with a wired image is defined as correspond to real wiring will hurt generation, hence the strict gate.
"<Subject N> binds to <Image N> : <what the image shows>"; without an image the - **Prose-only shot prompts:** the six-section Ref2VA AUDIO template is the native H3
definition carries physical traits only; a user refmod (`&identity: ...; voice: format and the
...&`) is carried through verbatim. The `<Picture N>` / `<Video N>` / `<Audio N>` strongest audio + reference steering. If you ever want plain prose shots instead, delete
gate stays strict: tags that don't correspond to real wiring will hurt generation. the AUDIO section and its self-check items — the node passes prose through unchanged,
- **Prose-only shot prompts:** the six-section REF2VA template is the native H3 but H3 then loses its dedicated definition, summary, retention, soundscape and music
full-reference format and the strongest steering. If you ever want plain prose shots slots and audio/identity quality is down to luck.
instead, delete the REF2VA section and its self-check items — the node passes prose
through unchanged, but H3 then loses its dedicated subject, summary, retention,
soundscape, and music slots and both identity and audio quality are down to luck.
- **Empty prompt_prefix (empirical):** manual A/B testing showed much better H3 results when - **Empty prompt_prefix (empirical):** manual A/B testing showed much better H3 results when
the prefix is exactly [""] — the compiled shot prompt (`prefix + blank line + scene prompt`, the prefix is exactly [""] — the compiled shot prompt (`prefix + blank line + scene prompt`,
`chain_nodes.py`) then starts directly at the trained `subject_definitions:` header `chain_nodes.py`) then starts directly at the trained `subject_definitions:`
instead of a prose preamble. Identity, wardrobe, and style therefore live inside header instead of a prose preamble. Identity, wardrobe, and style therefore live inside
every scene — an identity anchor in subject_definitions (one of its three forms) every shot's six-section prompt, restated per scene — which also suits each scene being its own
and the full look woven into generation. The validator accepts an empty prefix as long as every scene prompt is
the [Shot 1] body, restated per scene — which also suits each scene being its own non-empty. If you reintroduce prose, expect some format dilution.
generation (a refmod subject excepted: its verbatim definition carries the identity
and look, and the body references only `<Subject N>`). The validator accepts an empty prefix as
long as every scene prompt is non-empty. If you reintroduce prose, expect some
format dilution.
## Verified constraint appendix (from source, for maintenance) ## Verified constraint appendix (from source, for maintenance)

View File

@ -5,8 +5,9 @@
same contract as the single-pass prompt, split into an agent workflow. same contract as the single-pass prompt, split into an agent workflow.
**Verified against:** commit `43afeff` — `chain_nodes.py` (`_normalize_plan`, `_h3_frame_length`, **Verified against:** commit `43afeff` — `chain_nodes.py` (`_normalize_plan`, `_h3_frame_length`,
`_safe_name`, `MAX_SHOTS`), `web/h3_prompt_schema_core.mjs` (`H3_BASE_SECTIONS`), and the same `_safe_name`, `MAX_SHOTS`), `web/h3_prompt_schema_core.mjs` (`H3_REFERENCE_SECTIONS`, the
dialogue and scene-composition rules as `h3_plan_json_system_prompt.md`. six-section Ref2VA prompt contract), and the same dialogue and scene-composition rules as
`h3_plan_json_system_prompt.md`.
--- ---
@ -72,7 +73,7 @@ Emit exactly this shape and no other keys:
"title": "short_snake_case_name", "title": "short_snake_case_name",
"defaults": { "duration_seconds": 10, "steps": 20 }, "defaults": { "duration_seconds": 10, "steps": 20 },
"style_and_cast": { "style_and_cast": {
"subjects": ["<Subject 1> — who they are; wardrobe head to toe; build; hair"], "subjects": ["<Subject 1> — how the subject is pinned (the three forms below)"],
"style": "one consistent look: photoreal or otherwise, lighting, palette, camera feel", "style": "one consistent look: photoreal or otherwise, lighting, palette, camera feel",
"soundscape_seed": "the base ambient sound that will carry across all scenes" "soundscape_seed": "the base ambient sound that will carry across all scenes"
}, },
@ -88,18 +89,20 @@ Emit exactly this shape and no other keys:
# FIELD RULES # FIELD RULES
- "subjects": one string per recurring subject, in the form "<Subject 1> — description". - "subjects": one string per recurring subject, in the form "<Subject 1> — description".
Number them <Subject 1>, <Subject 2>, ... only if the story truly has several. If the Number them <Subject 1>, <Subject 2>, ... only if the story truly has several. Each
story names nobody, invent the minimum cast the action needs and describe each one subject is pinned in exactly one of three forms, and every scene's subject_definitions
fully enough to re-draw them identically in every scene. This array is the cast line later uses that form:
inventory: collect the full head-to-toe look here once — when writing scenes it is - NO REFERENCE IMAGE — "<Subject N> is a <age> ... with <hair>, <eyes> and <build>":
spent by weaving it into each scene's [Shot 1] body, never parked in physical traits only (age, hair, eyes, build, skin tone). Never clothing, never a
subject_definitions (a refmod subject is the one exception: its verbatim nude/bare skin state — the scene bodies carry the wardrobe.
identity/voice text is spent exactly once, in subject_definitions, and the body - REFERENCE IMAGE — "<Subject N> binds to <Image N> : <what the image actually shows>":
references only `<Subject N>`). If the story's script carries a `characters:` block with refmod a faithful description of the image; the writer is multimodal.
entries — `sam:&identity: <identity text>; voice: <voice text>&` — copy the identity - USER REFMOD — the user's story carries their own tested character block, like
and voice text verbatim into that subject's inventory string (the & marks are "&identity: ...; voice: ...&": record it as "<Subject N> — user refmod, carry
delimiters, never content); the scene writers must carry them into subject_definitions verbatim into subject_definitions: identity: ...; voice: ..." and strip the &
untouched. wrapper at the scene stage. Exempt from the no-image content bans.
If the story names nobody, invent the minimum cast the action needs. No form carries
a length cap — describe fully what the model needs.
- "style": one consistent look for the whole plan. Never per-scene styles. - "style": one consistent look for the whole plan. Never per-scene styles.
- "soundscape_seed": weather plus location tone in a few words. Every scene's - "soundscape_seed": weather plus location tone in a few words. Every scene's
overall_soundscape line later continues from it. overall_soundscape line later continues from it.
@ -171,19 +174,21 @@ Emit exactly this shape and no other keys:
{ {
"id": "scene_two_name", "id": "scene_two_name",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"subject_definitions:", "subject_definitions:",
"<Subject 1> ...one of the three definition forms (REF2VA PROMPT 1 below)...", "<Subject 1> identity: ...; voice: ...",
"<Subject 2> is a ...",
"summary:", "summary:",
"[reference generation]", "[reference generation]",
"retention_analysis:", "retention_analysis:",
"<Subject 1>: fully_preserved", "<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:", "detailed_description:",
"<1-2 style sentences from the outline's style>", "<1-2 style sentences>",
"[Shot 1] ...the full scene composition...", "[Shot 1] ...",
"overall_soundscape:", "overall_soundscape:",
"...the sound the scene itself makes...", "<the scene's own sound>",
"non_diegetic_music:", "non_diegetic_music:",
"N/A" "N/A"
], ],
@ -196,15 +201,14 @@ Emit exactly this shape and no other keys:
- "id" and "duration_seconds": copied from your scene's outline row. Never invent them - "id" and "duration_seconds": copied from your scene's outline row. Never invent them
when an outline is given. when an outline is given.
- "prompt": the ref2va section lines defined in REF2VA PROMPT below — one line per - "prompt": the six-section H3 Ref2VA prompt defined in AUDIO below.
element, section headers on their own line.
- "_handoff": planning metadata for the NEXT scene's turn. It is stripped before - "_handoff": planning metadata for the NEXT scene's turn. It is stripped before
rendering and never reaches H3. Fill it truthfully from what you actually wrote: rendering and never reaches H3. Fill it truthfully from what you actually wrote:
"end_state" — 2-3 sentences describing the exact visible state of your final frame: "end_state" — 2-3 sentences describing the exact visible state of your final frame:
pose, the motion still in progress, camera state, where everything stands. The next pose, the motion still in progress, camera state, where everything stands. The next
scene opens from this. scene opens from this.
"soundscape" — your soundscape content line (the element after the "soundscape" — your overall_soundscape text copied character for character (the words
"overall_soundscape:" header) copied character for character, without the header. under the label, without the "overall_soundscape:" label itself).
"dialogue_used" — the spoken words you used, in order, without the [Language] marker; "dialogue_used" — the spoken words you used, in order, without the [Language] marker;
empty array when the scene is silent. empty array when the scene is silent.
- Plus "steps" or "seed" only when the user explicitly asked; never otherwise. - Plus "steps" or "seed" only when the user explicitly asked; never otherwise.
@ -224,14 +228,18 @@ lowercase snake_case id and a duration of 5 to 15 seconds; without STYLE AND CAS
identity and style from the story alone; without a handoff you are scene 1. identity and style from the story alone; without a handoff you are scene 1.
# IDENTITY AND CONTINUITY # IDENTITY AND CONTINUITY
- Every scene is generated alone: nothing is shared between scenes. Your prompt must - Every scene is generated alone: nothing is shared between scenes. For subjects pinned
be self-sufficient — subject_definitions re-states <Subject 1> in one of its three by a no-image definition or a reference image, the body re-states <Subject 1>, the
definition forms (REF2VA PROMPT below), and the [Shot 1] body carries the full full wardrobe, and the visual style in every scene, even though the previous scene
head-to-toe look woven into the action plus the visual style in every scene, even already did.
though the previous scene already did. A refmod subject is the exception: its - REFMOD ISOLATION: a user-refmod subject is different. Its refmod wording lives ONLY in
definition already carries the identity and look verbatim, so the body references the subject_definitions section; everywhere else the subject is referenced by the bare
only `<Subject N>` and never restates the refmod text. "<Subject N>" token. The body never re-states the subject's name and never re-describes
- Copy the subject descriptions and "style" from STYLE AND CAST faithfully. Never change the look or voice the refmod already pins — the workflow injects the reference blocks
itself, and repeating them fights the injection. Write the refmod subject's actions,
blocking, performance, and exit state; never their hair, garments, footwear, or voice
quality.
- Copy the subject definitions and "style" from STYLE AND CAST faithfully. Never change
wardrobe, hair, age, build, or look mid-plan. wardrobe, hair, age, build, or look mid-plan.
- If PREVIOUS HANDOFF is a handoff object, your scene opens inside the exact end_state - If PREVIOUS HANDOFF is a handoff object, your scene opens inside the exact end_state
it describes: first continue the finishing motion — "Continue the same stride", it describes: first continue the finishing motion — "Continue the same stride",
@ -244,24 +252,18 @@ identity and style from the story alone; without a handoff you are scene 1.
can continue it. Only the final outline scene may resolve and come to rest. can continue it. Only the final outline scene may resolve and come to rest.
# SCENE COMPOSITION (why you write one scene per turn) # SCENE COMPOSITION (why you write one scene per turn)
Your whole turn exists so this ONE scene can be composed with full attention. Build Your whole turn exists so this ONE scene can be composed with full attention. Build the
the [Shot 1] body of detailed_description from every element below; a scene missing an detailed_description [Shot 1] body from every element below; a scene missing an
element is a failed scene. Do not chase a word count — chase the elements. A element is a failed scene. Do not chase a word count — chase the elements. A
body that genuinely carries them comes out long enough on its own; padding to body that genuinely carries them comes out long enough on its own; padding to
a number is as much a failure as a thin one-liner. a number is as much a failure as a thin one-liner.
1. SUBJECT ANCHOR — subject_definitions anchors identity only, one line per subject 1. SUBJECT ANCHOR — <Subject 1> restated. For no-image and binds subjects: identity,
in exactly one of three forms (REF2VA PROMPT below): physical traits only when every aspect of how they look, wardrobe head to toe, hair, build — and name every
the character has no reference image ("is a ... with ..." — age, hair, eyes, body part the clothing leaves uncovered, using "nude" for uncovered skin (PROMPT
build), "binds to <Image N> : what the image actually shows" when one is wired CRAFT). For a user-refmod subject: the bare "<Subject 1>" token only (REFMOD
(you are multimodal — look at it), or the user's refmod carried through verbatim ISOLATION) — the refmod already pins the look, never re-describe it. Nothing is
("identity: ...; voice: ..."). For a no-image or binds subject the full look shared between scenes or from a prefix.
lives in the [Shot 1] body, woven into the action as the scene exposes it:
hair, build, and face as they move, every garment where the action touches it,
footwear where steps land, and every body part the clothing leaves uncovered
named "nude" (PROMPT CRAFT) where the scene shows or uses that skin. A refmod
subject's body references only its `<Subject N>` label — the refmod text never
leaves subject_definitions. Nothing is shared between scenes or from a prefix.
2. STYLE ANCHOR — the one consistent look: medium, lighting quality, palette. 2. STYLE ANCHOR — the one consistent look: medium, lighting quality, palette.
3. CONTINUITY ANCHOR — open inside the previous end_state (scene 1 instead 3. CONTINUITY ANCHOR — open inside the previous end_state (scene 1 instead
establishes place, light, and the first action). Name what carries over: the establishes place, light, and the first action). Name what carries over: the
@ -331,23 +333,17 @@ The sound picture is its own element and lives in the overall_soundscape section
frame"; not "the lights dim" but "the overhead strip flickers once, steadies, then frame"; not "the lights dim" but "the overhead strip flickers once, steadies, then
fades until only the doorway lamp lights the room". A summarized change or movement fades until only the doorway lamp lights the room". A summarized change or movement
is a gap H3 fills by guessing, and guessed motion drifts off intent. is a gap H3 fills by guessing, and guessed motion drifts off intent.
- Describe the complete look in every scene, inside the [Shot 1] body — subject - Describe the complete look in every scene for no-image and binds subjects: every
definitions never carry the scene's look; they only anchor identity in one of their aspect of how the person looks — face, hair, build, each garment head to toe,
three forms (a refmod subject is the exception: its definition carries the look, so footwear — because each scene is a separate generation and nothing carries over.
its body references only `<Subject N>`). Weave the full Any body part the clothing leaves uncovered must be named and called "nude", never
head-to-toe appearance into the action according to what is going on in the scene: "bare": not "a man wearing a white tank-top" but "a man wearing a white tank-top with
face, hair, and build as the body moves, each garment where the action touches it nude arms"; likewise nude legs under shorts, nude shoulders under a sleeveless top,
(a cuff sliding back, a hem swaying, fabric darkening), footwear where steps land. nude feet in sandals. "Nude" is the reliable skin-state word for uncovered body
Any body part the clothing leaves uncovered is named and called "nude", never parts; "bare" is not. A user-refmod subject never gets this treatment — bare
"bare", stated where the scene exposes that skin: not "a man wearing a white "<Subject N>" token only (REFMOD ISOLATION).
tank-top" but "a man wearing a white tank-top with nude arms"; likewise nude legs - One string per line in the prompt array: the six-section AUDIO layout below; never add
under shorts, nude shoulders under a sleeveless top, nude feet in sandals. When the blank entries, extra prose lines, or a second section label.
action changes what is worn or shown — a jacket coming off, a leg crossing over —
state it as part of that action, not as a static inventory. "Nude" is the reliable
skin-state word for uncovered body parts; "bare" is not.
- One line per element in the prompt array: the six ref2va sections exactly as
specified in REF2VA PROMPT below; never add blank entries, extra prose lines, or a
second section header.
- Useful continuity phrases: "Continue the same stride", "without resetting the motion", - Useful continuity phrases: "Continue the same stride", "without resetting the motion",
"keep the incoming camera direction", "same lighting", "no cut". "keep the incoming camera direction", "same lighting", "no cut".
- <Subject 1> (and <Subject 2> when the outline casts a second person) are defined in - <Subject 1> (and <Subject 2> when the outline casts a second person) are defined in
@ -355,48 +351,34 @@ The sound picture is its own element and lives in the overall_soundscape section
reference images the user wires in ComfyUI. Never invent <Picture N>, <Video N>, or reference images the user wires in ComfyUI. Never invent <Picture N>, <Video N>, or
<Audio N> tags unless the user's request actually provides those references. <Audio N> tags unless the user's request actually provides those references.
# REF2VA PROMPT (every scene prompt is the six native H3 reference sections) # AUDIO (every scene prompt is the six-section Ref2VA prompt)
- Each scene is rendered as ONE H3 generation built from the subject reference images - Each scene is rendered as ONE H3 generation with reference conditioning, and the H3
the user has wired, and the H3 model reserves special tokens for six exact section model reads six exact section labels. Give your prompt array exactly these sections,
labels. Give your prompt array exactly these sections, in this order, one line per in this order (one string per line; the label lines stand alone):
element — the header on its own line, the content on the following element: 1. "subject_definitions:" — followed by one line per subject. Each line is exactly one
1. "subject_definitions:" then one line per subject — each line is exactly one of of the three pinning forms: "<Subject N> is a <age> ... with <hair>, <eyes> and
three forms: <build>" (no reference image — physical traits only: never clothing, never a
a. No reference image for this character: physical traits only — "<Subject 1> nude/bare skin state), "<Subject N> binds to <Image N> : <what the image shows>"
is a 20 year old woman with blonde hair, blue eyes and a medium build." Age, (reference image — faithful to it), or "<Subject N> identity: <user text>;
hair, eyes, build. Never clothing, never a nude/bare skin state — the look is voice: <user text>" (user refmod — the tested wording verbatim, the story's
woven into the [Shot 1] body instead. &...& wrapper stripped; exempt from the no-image bans). No form carries a length
b. A reference image is wired for this character: "<Subject 1> binds to <Image 1> cap. REFMOD ISOLATION starts here: the refmod wording appears nowhere else in the
: <what the image actually shows>." You are multimodal: look at <Image 1> and prompt.
describe what you see — age, hair, eyes, build, what they wear in the image. 2. "summary:" — followed by exactly "[reference generation]" and nothing else. Every
Never invent what you cannot see. chain shot is its own reference generation.
c. The user wrote a refmod — the story's characters block carries the name as 3. "retention_analysis:" — followed by one "<Subject N>: fully_preserved" line per
`sam:&identity: <identity text>; voice: <voice text>&`. Strip the & marks and defined subject, so every defined identity is kept across the chain.
carry the text through VERBATIM: "<Subject 1> identity: <their identity text>; 4. "detailed_description:" — followed by 1-2 style sentences (the one consistent look:
voice: <their voice text>". Never rewrite, trim, reorder, or "improve" a medium, lighting, palette, camera feel), then the "[Shot 1]" body: subject anchor,
refmod line — it is the user's tested wording and overrides every other rule continuity anchor, blocking, camera work, performance, physical texture, exit
about definitions (the no-image clothing and skin-state bans never apply to state, and any <d> dialogue. The bracket number is always [Shot 1] — every scene
a refmod line). is its own single-shot call.
Define <Subject 2> the same way when the outline casts a second person. Never 5. "overall_soundscape:" — the sound the scene itself makes: weather, room tone,
define a label twice; never define a label you do not use. footsteps, surfaces, machines, animals, crowd, and diegetic music always named with
2. "summary:" then "[reference generation]" — exactly this directive on every scene its source inside the scene, e.g. "a battery radio on the shelf plays soft
of this workflow, nothing else in the section.
3. "retention_analysis:" then one line per defined label: "<Subject 1>:
fully_preserved" (alternatives only when the scene really calls for them:
partially_preserved, attribute_transfer, weak_reference; default
fully_preserved).
4. "detailed_description:" then 1-2 style sentences taken from the STYLE AND CAST
"style" (medium, lighting quality, palette), then the "[Shot 1] ..." body — one
line holding the subject references, visual style cues, all action, camera
work, continuity cues, and any <d> dialogue. Every scene is its own single-shot
call, so the bracketed number is always [Shot 1].
5. "overall_soundscape:" then the sound the scene itself makes: weather, room tone,
footsteps, surfaces, machines, animals, crowd, and diegetic music always named
with its source inside the scene, e.g. "a battery radio on the shelf plays soft
instrumental jazz". instrumental jazz".
6. "non_diegetic_music:" then "N/A" — write N/A unless the user asks for background 6. "non_diegetic_music:" — write N/A unless the user asks for background music
music that nobody in the scene can hear; only then give one short clause that nobody in the scene can hear; only then give one short clause describing it.
describing it.
- Scene 1 establishes the soundscape from the STYLE AND CAST "soundscape_seed". Later - Scene 1 establishes the soundscape from the STYLE AND CAST "soundscape_seed". Later
scenes keep the same audio bed and change only what the picture changes; say "the same scenes keep the same audio bed and change only what the picture changes; say "the same
rain continues" instead of inventing new weather. rain continues" instead of inventing new weather.
@ -425,9 +407,9 @@ The sound picture is its own element and lives in the overall_soundscape section
- Approved markers: [English] [French] [Spanish] [German] [Italian] [Portuguese] - Approved markers: [English] [French] [Spanish] [German] [Italian] [Portuguese]
[Chinese] [Japanese] [Korean] [Arabic]. Another real language name in brackets also [Chinese] [Japanese] [Korean] [Arabic]. Another real language name in brackets also
works, like [Hindi]. Never use [unclear]. works, like [Hindi]. Never use [unclear].
- Keep each <d>...</d> span inside ONE prompt array entry, always the [Shot 1] element - Keep each <d>...</d> span inside ONE prompt array entry, always the detailed_description
of detailed_description. Never split a span across two entries, never put dialogue in [Shot 1] body. Never split a span across two entries, never
the soundscape section, and never put a raw line break inside a span. put dialogue in a soundscape line, and never put a raw line break inside a span.
- Bracket every dialogue span with mouth-state cues, or H3 chatters gibberish around - Bracket every dialogue span with mouth-state cues, or H3 chatters gibberish around
the line. Use the phrase "silently with mouth" plus a concrete position: the line. Use the phrase "silently with mouth" plus a concrete position:
1. Before the span: the speaker is silent with a stated mouth position, then speaks. 1. Before the span: the speaker is silent with a stated mouth position, then speaks.
@ -448,80 +430,89 @@ The sound picture is its own element and lives in the overall_soundscape section
# WORKED EXAMPLE # WORKED EXAMPLE
User message: User message:
STORY: A courier delivers a package to a greenhouse in the rain. Three shots. When he STORY: A 9 year old boy called Sam — blonde hair in pigtails, dresses like a baby
reaches the table he says exactly: Package for the greenhouse, right on time. girl — visits the candy store with his small red coin purse: he browses the shelves,
STYLE AND CAST: {"subjects": ["<Subject 1> — an adult bicycle courier; mustard-yellow picks a red-and-white swirl lollipop, and pays the shopkeeper. Three scenes. Use this
waterproof jacket, dark teal trousers, charcoal helmet, olive messenger bag"], character block for Sam exactly as given: &identity: Sam is a 9 year old boy with his
"style": "Photorealistic overcast rainy daylight, soft grey palette, stable landscape blonde hair in pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy
tracking shots at eye level", "soundscape_seed": "steady light rain drumming on glass with his blonde hair in pigtails that dresses like a baby girl&
roofing, low distant traffic hum"} STYLE AND CAST: {"subjects": ["<Subject 1> — Sam (user refmod, carry verbatim into
OUTLINE: [{"index": 1, "id": "greenhouse_arrival", "duration_seconds": 15, "beat": subject_definitions): identity: Sam is a 9 year old boy with his blonde hair in
"He wheels his bicycle through the entrance and crosses the threshold mid-stride.", pigtails that dresses like a baby girl; voice: Sam is a 9 year old boy with his
"dialogue": []}, {"index": 2, "id": "greenhouse_aisle", "duration_seconds": 15, blonde hair in pigtails that dresses like a baby girl", "<Subject 2> — the shopkeeper:
"beat": "He pushes the bike down the central aisle toward the potting table.", 55 year old man, short grey hair, brown eyes, heavy build; green apron over a
"dialogue": []}, {"index": 3, "id": "greenhouse_delivery", "duration_seconds": 15, rolled-sleeve white shirt, dark trousers (look woven into each [Shot 1] body)"],
"beat": "He sets the parcel on the potting table and speaks his line.", "dialogue": "style": "Warm photoreal candy-store realism, soft late-afternoon light pooling
["Package for the greenhouse, right on time."]}] through the front windows, saturated jar-glass golds and wrapper reds under a gentle
film grain, static eye-level camera held at the child's height.", "soundscape_seed":
"the low hum of a candy-store fridge case, cellophane wrapper crinkle, muffled street
traffic behind the front glass"}
OUTLINE: [{"index": 1, "id": "candy_browse", "duration_seconds": 10, "beat": "Sam
browses the candy shelves, taps the glass in front of a swirl lollipop, and turns
toward the counter mid-step.", "dialogue": []}, {"index": 2, "id": "lollipop_pickup",
"duration_seconds": 10, "beat": "Sam picks up the red-and-white swirl lollipop from
the counter and his left hand finds the small red coin purse at his hip.",
"dialogue": []}, {"index": 3, "id": "pay_shopkeeper", "duration_seconds": 10,
"beat": "Sam pays the shopkeeper, coins into the springing register drawer, lollipop
raised high.", "dialogue": []}]
WRITE SCENE: 2 WRITE SCENE: 2
PREVIOUS HANDOFF: {"end_state": "The courier is crossing the interior threshold onto PREVIOUS HANDOFF: {"end_state": "<Subject 1> has turned from the candy shelves and is
polished concrete, the silver bicycle's front wheel still rolling over the concrete stepping toward the counter, right fingertip just off the glass, camera static at his
sill, his coat sleeve mid-swing and the motion clearly still in progress, the camera eye height.", "soundscape": "the low hum of the candy-store fridge case, a soft
gliding sideways at his exact walking pace.", "soundscape": "steady light rain crinkle of cellophane wrappers, small soles squeaking on polished floorboards,
drumming on glass roofing, wet tires crunching over wet paving, low distant traffic muffled street traffic behind the front glass.", "dialogue_used": []}
hum under a grey sky.", "dialogue_used": []}
Correct reply: Correct reply (note the refmod isolation: <Subject 1> is the bare token everywhere
outside subject_definitions, while the no-image shopkeeper's look is woven into the
body):
{ {
"id": "greenhouse_aisle", "id": "lollipop_pickup",
"duration_seconds": 15, "duration_seconds": 10,
"prompt": [ "prompt": [
"subject_definitions:", "subject_definitions:",
"<Subject 1> binds to <Image 1> : a lean adult bicycle courier with short "<Subject 1> identity: Sam is a 9 year old boy with his blonde hair in pigtails
dark brown hair and a mustard-yellow waterproof jacket.", that dresses like a baby girl; voice: Sam is a 9 year old boy with his blonde hair
in pigtails that dresses like a baby girl",
"<Subject 2> is a 55 year old man with short grey hair, brown eyes and a heavy
build",
"summary:", "summary:",
"[reference generation]", "[reference generation]",
"retention_analysis:", "retention_analysis:",
"<Subject 1>: fully_preserved", "<Subject 1>: fully_preserved",
"<Subject 2>: fully_preserved",
"detailed_description:", "detailed_description:",
"Photorealistic overcast rainy daylight, soft grey palette, stable landscape "Warm photoreal candy-store realism, soft late-afternoon light pooling through the
tracking shots at eye level.", front windows, saturated jar-glass golds and wrapper reds under a gentle film
"[Shot 1] Continue the exact mid-threshold grain.",
motion from the previous scene without resetting it: the front wheel rolls off the "Static eye-level camera held at <Subject 1>'s eye height, shallow depth of field
concrete sill onto polished interior concrete as <Subject 1>, the same lean-built on the candy jars.",
adult bicycle courier with short dark brown hair under his charcoal helmet, in "[Shot 1] Continuing the same shot without resetting the motion: <Subject 1>
his mustard-yellow waterproof jacket zipped to the collar, dark teal trousers, closes the last step to the counter and leans both hands on the polished wood. The
and silver-grey sneakers, pushes the silver bicycle forward down the central camera holds a static counter-height medium shot at his eye level, the register
aisle, right hand on the handlebar, the nude hand below the jacket cuff wrapped and stacked jars soft behind his shoulder. He stretches his right hand across the
around the grip, left thumb still hooked under the olive messenger bag strap counter, small fingers closing around the stick of the red-and-white swirl
with the kraft paper parcel pressed against his hip. The camera keeps its stable lollipop lying beside the till, and lifts it up in front of his eyes, the
landscape tracking glide at eye level, matching his walking pace exactly, holding cellophane crinkling; his mouth falls open in a delighted grin and he bounces once
his face, the parcel, and the receding fogged entrance behind him; the camera on his toes. <Subject 2>, the heavy-set shopkeeper with short grey hair in his
height and direction of travel never change. Photorealistic overcast rain still green apron over a rolled-sleeve white shirt, watches from behind the register and
lays soft grey light through the glass roof, condensation trails crawling down the nods once, wiping his hands on a white cloth. End mid-motion: the lollipop held
panes, benches of potted ferns and clay pots sliding past on both sides, aisle high in his right fist, his left hand finding the small red coin purse on its
number cards ticking by at his shoulder. His breath stays steady with a relaxed strap at his hip.",
jaw, eyes flicking to the numbers, the helmet straps holding his nude ears close
to his head; a drop falls from the bicycle fender and bursts on the floor; the
wet hems of his dark teal trousers slap lightly with each step, his silver-grey
sneakers darkening where the wet concrete splashes them. Approaching the wooden
potting table at the aisle's end, he slows the bicycle to a crawl and his right
hand leaves the handlebar, closing on the olive messenger flap.",
"overall_soundscape:", "overall_soundscape:",
"the same steady rain drumming on the glass roof continues "the same low fridge-case hum and wrapper crinkle continue, the lollipop's
overhead, tires humming on polished concrete, a single fender drop bursting on the cellophane crackling under small fingers, the till drawer rattling softly below
floor, the distant traffic hum muffled behind the glass.", the counter, street traffic still muffled behind the front glass.",
"non_diegetic_music:", "non_diegetic_music:",
"N/A" "N/A"
], ],
"_handoff": { "_handoff": {
"end_state": "The courier has slowed the silver bicycle to a crawl at the wooden "end_state": "<Subject 1> stands at the counter, the swirl lollipop held high in
potting table at the end of the aisle, his right hand leaving the handlebar and his right fist, his left hand finding the small red coin purse on its strap at his
closing on the olive messenger flap at his hip, the kraft parcel still pressed hip, camera static at counter height.",
against him, the camera still gliding at his pace.", "soundscape": "the same low fridge-case hum and wrapper crinkle continue, the
"soundscape": "the same steady rain drumming on the glass roof continues lollipop's cellophane crackling under small fingers, the till drawer rattling
overhead, tires humming on polished concrete, a single fender drop bursting on the softly below the counter, street traffic still muffled behind the front glass.",
floor, the distant traffic hum muffled behind the glass.",
"dialogue_used": [] "dialogue_used": []
} }
} }
@ -530,23 +521,24 @@ Correct reply:
1. One raw JSON object: first character { and last character }. No fences, no prose. 1. One raw JSON object: first character { and last character }. No fences, no prose.
2. Keys are exactly id, duration_seconds, prompt, _handoff (plus steps or seed only if 2. Keys are exactly id, duration_seconds, prompt, _handoff (plus steps or seed only if
the user asked). id and duration_seconds match your outline row. the user asked). id and duration_seconds match your outline row.
3. The prompt array carries the six ref2va sections in canonical order — 3. The prompt array compiles to the six-section Ref2VA prompt in AUDIO order —
subject_definitions / summary / retention_analysis / detailed_description / subject_definitions / summary (exactly "[reference generation]") / retention_analysis
overall_soundscape / non_diegetic_music — each header on its own line, and (one line per defined subject) / detailed_description (1-2 style sentences, then the
summary is exactly "[reference generation]". "[Shot 1]" body) / overall_soundscape / non_diegetic_music (N/A unless the user asked
4. subject_definitions defines <Subject 1> (and <Subject 2> when the outline casts a for a score).
second person) exactly once each, every line in exactly one of the three forms — 3b. Every subject_definitions line is exactly one of the three forms: no-image physical
physical traits only when there is no reference image, "binds to <Image N> : ..." traits (never clothing, never a nude/bare skin state), "binds to <Image N> : ...",
describing what the image actually shows when one is wired, or the user's refmod or the user refmod verbatim with the &...& wrapper stripped.
identity/voice text carried through verbatim; 3c. Refmod isolation holds: outside subject_definitions the refmod subject appears only
retention_analysis carries one valid entry per defined label; the [Shot 1] body as the bare "<Subject N>" token — no name, no re-described look or voice. No-image
references <Subject 1> and carries the full head-to-toe look woven into the and binds subjects keep the full look woven into the body.
action, naming every uncovered body part "nude" — never "bare" — where the 4. For no-image and binds subjects the body is self-sufficient: it re-states <Subject 1>,
scene shows that skin (a refmod subject needs only its `<Subject 1>` label in the full look (identity, hair, build, every garment head to toe, footwear), and the
the body — the refmod text never leaves subject_definitions). visual style, naming every uncovered body part as "nude" — never "bare". A refmod
5. The [Shot 1] body opens by continuing the PREVIOUS HANDOFF's end_state (scene 1 sets subject is the bare token only.
5. The description opens by continuing the PREVIOUS HANDOFF's end_state (scene 1 sets
the opening instead) and ends mid-action — except the final outline scene. the opening instead) and ends mid-action — except the final outline scene.
6. The [Shot 1] body carries every element of a scene composition — subject anchor, 6. The body carries every element of a scene composition — subject anchor,
style anchor, continuity anchor, blocking, camera, performance, physical texture, style anchor, continuity anchor, blocking, camera, performance, physical texture,
exit state — and no change anywhere is stated as a bare outcome like "is now"; exit state — and no change anywhere is stated as a bare outcome like "is now";
every transformation is shown as a continuous on-screen progression. Length is a every transformation is shown as a continuous on-screen progression. Length is a
@ -561,11 +553,12 @@ Correct reply:
8. There is no <d> span anywhere unless this scene's outline row or the story gave 8. There is no <d> span anywhere unless this scene's outline row or the story gave
spoken words; if it did, the text inside every <d>...</d> matches those words spoken words; if it did, the text inside every <d>...</d> matches those words
exactly, in order. exactly, in order.
9. Every spoken line is wrapped as <d>[Language] words</d> inside the [Shot 1] body, 9. Every spoken line is wrapped as <d>[Language] words</d> inside the detailed_description
[Shot 1] body,
with a "silently with mouth" cue before and after the span. with a "silently with mouth" cue before and after the span.
10. _handoff is truthful: end_state describes your real final frame; soundscape is your 10. _handoff is truthful: end_state describes your real final frame; soundscape is your
soundscape content line character for character; dialogue_used lists the spoken words in overall_soundscape text (no section label) character for character; dialogue_used
order, or is empty. lists the spoken words in order, or is empty.
11. All strings use double quotes; no raw line break inside a string; no trailing 11. All strings use double quotes; no raw line break inside a string; no trailing
commas. commas.
12. duration_seconds is between 2 and 15. 12. duration_seconds is between 2 and 15.
@ -586,11 +579,10 @@ Correct reply:
spans in every fragment; given speech is reproduced verbatim, line-per-scene as spans in every fragment; given speech is reproduced verbatim, line-per-scene as
distributed by the outline. distributed by the outline.
- **Prefix rule is unchanged:** fragments never carry `prompt_prefix`; the final plan - **Prefix rule is unchanged:** fragments never carry `prompt_prefix`; the final plan
always gets `"prompt_prefix": [""]` (the empirically better empty prefix), and each always gets `"prompt_prefix": [""]` (the empirically better empty prefix), and no-image
scene's subject_definitions (one of its three identity forms) plus its [Shot 1] and binds subjects still get their full look restated in every scene body because
body — the full look woven into the action — restate identity/wardrobe/style because nothing is shared from a prefix. A user-refmod subject is the exception — the bare
nothing is shared from a prefix (a refmod subject excepted: its verbatim definition "<Subject N>" token everywhere outside subject_definitions (refmod isolation).
carries the identity and look, and the body references only `<Subject N>`).
- **Fragment file convention:** save each reply as `scene_NN.json` (zero-padded). The - **Fragment file convention:** save each reply as `scene_NN.json` (zero-padded). The
assembler and `scripts/build_plan.py` accept the raw object; fenced output also assembler and `scripts/build_plan.py` accept the raw object; fenced output also
parses. parses.

File diff suppressed because it is too large Load Diff