Some checks are pending
Publish to Comfy registry / Publish Custom Node to registry (push) Waiting to run
146 lines
4.6 KiB
Python
146 lines
4.6 KiB
Python
from pathlib import Path
|
|
import sys
|
|
|
|
import torch
|
|
|
|
|
|
def _load_audio_utils():
|
|
sys.path.insert(0, str(Path(__file__).parents[1]))
|
|
from utils import audio as audio_utils
|
|
|
|
return audio_utils
|
|
|
|
|
|
def _audio(values: list[float], sample_rate: int = 4) -> dict:
|
|
return {
|
|
"waveform": torch.tensor(values, dtype=torch.float32).reshape(1, 1, -1),
|
|
"sample_rate": sample_rate,
|
|
}
|
|
|
|
|
|
def test_iter_valid_audio_inputs_flattens_lists_and_ignores_empty_items():
|
|
audio_utils = _load_audio_utils()
|
|
first = _audio([1, 2])
|
|
second = _audio([3, 4])
|
|
|
|
result = audio_utils.iter_valid_audio_inputs(
|
|
{
|
|
"audio_0": [None, [], first],
|
|
"audio_1": None,
|
|
"audio_2": [[second], {"waveform": None}],
|
|
}
|
|
)
|
|
|
|
assert result == [first, second]
|
|
|
|
|
|
def test_audio_is_silent_detects_empty_and_nonzero_waveforms():
|
|
audio_utils = _load_audio_utils()
|
|
|
|
assert audio_utils.audio_is_silent(_audio([0.0, 0.0])) is True
|
|
assert audio_utils.audio_is_silent(_audio([0.0, 0.01])) is False
|
|
assert audio_utils.audio_is_silent(
|
|
{"waveform": torch.empty(1, 1, 0), "sample_rate": 4}
|
|
) is True
|
|
|
|
|
|
def test_merge_audio_inputs_merges_three_tracks_with_padding_and_normalization():
|
|
audio_utils = _load_audio_utils()
|
|
|
|
result = audio_utils.merge_audio_inputs(
|
|
[
|
|
_audio([0.5, 0.5, 0.5, 0.5]),
|
|
_audio([0.5, 0.5]),
|
|
_audio([1.0, 1.0, 1.0, 1.0]),
|
|
],
|
|
"add",
|
|
)
|
|
|
|
assert result["sample_rate"] == 4
|
|
assert result["waveform"].shape == (1, 1, 4)
|
|
assert result["waveform"].flatten().tolist() == [1.0, 1.0, 0.75, 0.75]
|
|
|
|
|
|
def test_merge_audio_inputs_returns_none_when_lists_have_no_audio():
|
|
audio_utils = _load_audio_utils()
|
|
|
|
assert audio_utils.merge_audio_inputs(audio_utils.iter_valid_audio_inputs([], None, [None])) is None
|
|
|
|
|
|
def test_merge_audio_inputs_resamples_to_higher_sample_rate(monkeypatch):
|
|
audio_utils = _load_audio_utils()
|
|
calls = []
|
|
|
|
def fake_resample(waveform, source_rate, target_rate):
|
|
calls.append((source_rate, target_rate))
|
|
repeat = target_rate // source_rate
|
|
return waveform.repeat_interleave(repeat, dim=-1)
|
|
|
|
monkeypatch.setitem(
|
|
sys.modules,
|
|
"torchaudio",
|
|
type("FakeTorchaudio", (), {"functional": type("Functional", (), {"resample": staticmethod(fake_resample)})}),
|
|
)
|
|
|
|
result = audio_utils.merge_audio_inputs([_audio([1.0, 1.0], 2), _audio([1.0, 1.0, 1.0, 1.0], 4)])
|
|
|
|
assert calls == [(2, 4)]
|
|
assert result["sample_rate"] == 4
|
|
assert result["waveform"].shape == (1, 1, 4)
|
|
|
|
|
|
def test_merge_audio_inputs_after_concatenates_in_order_and_converts_mono_to_stereo():
|
|
audio_utils = _load_audio_utils()
|
|
|
|
result = audio_utils.merge_audio_inputs([_audio([1.0, 2.0]), _audio([3.0])], "after")
|
|
|
|
assert result["sample_rate"] == 4
|
|
assert result["waveform"].shape == (1, 2, 3)
|
|
assert result["waveform"][0, 0].tolist() == [1.0, 2.0, 3.0]
|
|
assert result["waveform"][0, 1].tolist() == [1.0, 2.0, 3.0]
|
|
|
|
|
|
def test_merge_audio_inputs_before_concatenates_in_reverse_order():
|
|
audio_utils = _load_audio_utils()
|
|
|
|
result = audio_utils.merge_audio_inputs([_audio([1.0]), _audio([2.0]), _audio([3.0])], "before")
|
|
|
|
assert result["sample_rate"] == 4
|
|
assert result["waveform"].shape == (1, 2, 3)
|
|
assert result["waveform"][0, 0].tolist() == [3.0, 2.0, 1.0]
|
|
|
|
|
|
def test_merge_audio_inputs_after_handles_six_autogrow_slots():
|
|
audio_utils = _load_audio_utils()
|
|
slots = {f"audio_{index}": _audio([float(index)]) for index in range(6)}
|
|
|
|
result = audio_utils.merge_audio_inputs(audio_utils.iter_valid_audio_inputs(slots), "after")
|
|
|
|
assert result["sample_rate"] == 4
|
|
assert result["waveform"].shape == (1, 2, 6)
|
|
assert result["waveform"][0, 0].tolist() == [0.0, 1.0, 2.0, 3.0, 4.0, 5.0]
|
|
|
|
|
|
def test_save_audio_to_temp_wav_serializes_first_batch(monkeypatch, tmp_path):
|
|
audio_utils = _load_audio_utils()
|
|
calls = {}
|
|
|
|
class FakeTorchaudio:
|
|
@staticmethod
|
|
def save(path, waveform, sample_rate):
|
|
calls["save"] = (path, waveform.clone(), sample_rate)
|
|
Path(path).write_bytes(b"wav")
|
|
|
|
monkeypatch.setitem(sys.modules, "torchaudio", FakeTorchaudio)
|
|
monkeypatch.setattr(audio_utils.folder_paths, "get_temp_directory", lambda: str(tmp_path))
|
|
audio = {
|
|
"waveform": torch.tensor([[[0.1, 0.2]], [[0.3, 0.4]]]),
|
|
"sample_rate": 16000,
|
|
}
|
|
|
|
output = audio_utils.save_audio_to_temp_wav(audio)
|
|
|
|
assert output is not None and output.is_file()
|
|
assert calls["save"][1].shape == (1, 2)
|
|
assert calls["save"][2] == 16000
|