Comfy-Org/ComfyUI · error · ValueError

Total reference audio duration is {total_audio_duration:.1f}

Error message

Total reference audio duration is {total_audio_duration:.1f}s. Maximum is {limits['max_total_seconds']} seconds.

What it means

The combined duration of all reference audios (sum of waveform-length/sample_rate across reference_audios inputs) must not exceed limits["max_total_seconds"] (30.1s on Seedance 2.5). The node computes the total locally and rejects the batch before any network call, quoting the measured total and the cap.

Source

Thrown at comfy_api_nodes/nodes_bytedance.py:2816

            except ValueError:
                raise
            except Exception:
                pass
        if total_video_duration > limits["max_total_seconds"]:
            raise ValueError(
                f"Total reference video duration is {total_video_duration:.1f}s. "
                f"Maximum is {limits['max_total_seconds']} seconds."
            )

        total_audio_duration = 0.0
        for i, key in enumerate(reference_audios, 1):
            audio = reference_audios[key]
            dur = int(audio["waveform"].shape[-1]) / int(audio["sample_rate"])
            if dur < 1.8:
                raise ValueError(f"Reference audio {i} is too short: {dur:.1f}s. Minimum duration is 1.8 seconds.")
            total_audio_duration += dur
        if total_audio_duration > limits["max_total_seconds"]:
            raise ValueError(
                f"Total reference audio duration is {total_audio_duration:.1f}s. "
                f"Maximum is {limits['max_total_seconds']} seconds."
            )

        asset_labels = _build_asset_labels(
            reference_assets,
            reference_image_assets,
            reference_video_assets,
            reference_audio_assets,
            len(reference_images),
            len(reference_videos),
            len(reference_audios),
        )
        prompt_text = _rewrite_asset_refs(model["prompt"], asset_labels)

        content: list[TaskTextContent | TaskImageContent | TaskVideoContent | TaskAudioContent] = [
            TaskTextContent(text=prompt_text),
        ]

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Trim audio references so their total stays under the stated maximum.
2. Keep one or two representative clips instead of many.

total = sum(int(a["waveform"].shape[-1]) / int(a["sample_rate"]) for a in reference_audios.values())
assert total <= seedance2_reference_limits(model_id)["max_total_seconds"]

When it happens

Trigger: Connecting multiple voice clips whose summed durations exceed max_total_seconds on the Seedance 2.5 reference node.

Common situations: Adding 'just one more' voice sample past the budget; referencing full spoken sentences when only a phrase is needed.

Related errors

• Reference audio {i} is too short: {dur:.1f}s. Minimum durati
• Too many reference audios: {total_audios} (audios={len(refer
• Reference video {i} is too short: {dur:.1f}s. Minimum durati
• Total reference video duration is {total_video_duration:.1f}
• The provider rejected the audio track this model generated f

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/4147b66cd20cae29. Report an issue: GitHub.