unslothai/unsloth · error · ValueError

{Path(path).name} carries no audio track. MiniMax-H3 denoise

Error message

{Path(path).name} carries no audio track. MiniMax-H3 denoises video and audio in one packed sequence, so its training clips must have sound.

What it means

Raised by decode_clip() when the container has a video track but no audio track. MiniMax-H3 denoises video and audio in one packed sequence, and the audio rows are part of the training objective, so an audio-less clip is refused rather than silently trained as silence (a silent target would teach the model to stop generating sound).

Source

Thrown at studio/backend/core/training/diffusion_h3_clips.py:294

    is the dataset contract -- pre-trim to the training duration -- but it used to be silent,
    which is how a caption describing a whole scene ended up on its first second. ``on_note``
    is called once per over-long clip with the numbers, so the run reports it.

    ``waveform`` is a float32 array of shape ``(2, h3_audio_sample_count(num_frames))`` at
    32 kHz. A mono source is duplicated to both channels; a clip with **no** audio track is
    refused rather than silently trained as silence, because the audio rows are in the objective
    and a silent target teaches the model to stop generating sound.
    """
    import av
    import numpy as np
    from PIL import Image

    target_samples = h3_audio_sample_count(num_frames)
    with av.open(str(path)) as container:
        if not container.streams.video:
            raise ValueError(f"{Path(path).name} carries no video track.")
        if not container.streams.audio:
            raise ValueError(
                f"{Path(path).name} carries no audio track. MiniMax-H3 denoises video and audio "
                f"in one packed sequence, so its training clips must have sound."
            )
        stream = container.streams.video[0]
        source_fps = float(stream.average_rate or stream.guessed_rate or H3_FPS) or float(H3_FPS)
        # Container duration, in seconds, for the over-long note below. Best effort: an unknown
        # duration simply means no note, never a failed decode.
        source_duration_s = 0.0
        try:
            if stream.duration is not None and stream.time_base is not None:
                source_duration_s = float(stream.duration * stream.time_base)
            elif getattr(container, "duration", None):
                source_duration_s = float(container.duration) / 1_000_000.0
        except Exception:  # noqa: BLE001 -- a note is not worth failing a decode over
            source_duration_s = 0.0

        frames: list[Any] = []
        next_target = 0

View on GitHub (pinned to 203007d190)

Solutions

1. Replace the clip with a version that has a real audio track, or add a soundtrack before training.
2. Exclude silent clips from the dataset via a pre-scan (container.streams.audio empty).
3. For generated clips, pair them with a licensed audio bed so captions describing sound have a target.

Example fix

# before
# dataset of AI-generated silent .mp4s -> "carries no audio track"

# after
# mux an audio bed: ffmpeg -loop 1 -i silent.mp4 -i bed.wav -shortest out.mp4
# or filter at ingest:
with av.open(p) as c:
    if not c.streams.audio:
        skip(p)  # report and exclude

import av

def has_audio_track(path: str) -> bool:
    try:
        with av.open(path) as c:
            return bool(c.streams.audio)
    except av.error.InvalidDataError:
        return False

try:
    frames, waveform = decode_clip(p, num_frames=n, width=w, height=h)
except ValueError as e:
    if "no audio track" in str(e):
        skip_and_log(p)
    else:
        raise

When it happens

Trigger: Passing a muted render, a screen recording captured without system audio, an exported animation/GIF-to-mp4 conversion, or stock b-roll delivered without a soundtrack.

Common situations: AI-generated video clips (many generators emit silent files); OBS/screen-capture defaults with audio disabled; stock footage proxies stripped of audio for size; timelapse exports.

Related errors

• {Path(path).name} carries {have_s:.2f}s of audio for a {want
• MiniMax-H3 was trained on aspect ratios from 1:4 to 4:1; thi
• {Path(path).name} carries no video track.
• {Path(path).name} decoded to {len(frames)} frames at {H3_FPS
• {Path(path).name} decoded to no audio samples.

AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15). Data as JSON: /api/errors/67274f86ddb5d754. Report an issue: GitHub.