Comfy-Org/ComfyUI · error · ValueError

Unsupported audio_channels: {audio_channels}

Error message

Unsupported audio_channels: {audio_channels}

What it means

The vocoder wrapper handles decoder output of 1 channel (squeeze to mono vocoder input) or 2 channels (stereo); decoder.out_ch anything else (e.g. 4 or 6 from an unsupported checkpoint) has no defined vocoder path and run_vocoder refuses it.

Source

Thrown at comfy/ldm/lightricks/vae/audio_vae.py:197

            target_length -= LATENT_DOWNSAMPLE_FACTOR - 1
        return (
            batch,
            self.autoencoder.decoder.out_ch,
            target_length,
            self.autoencoder.mel_bins,
        )

    def num_of_latents_from_frames(self, frames_number: int, frame_rate: float) -> int:
        return round((float(frames_number) / frame_rate) * self.latents_per_second)

    def run_vocoder(self, mel_spec: torch.Tensor) -> torch.Tensor:
        audio_channels = self.autoencoder.decoder.out_ch
        vocoder_input = mel_spec.transpose(2, 3)

        if audio_channels == 1:
            vocoder_input = vocoder_input.squeeze(1)
        elif audio_channels != 2:
            raise ValueError(f"Unsupported audio_channels: {audio_channels}")

        return self.vocoder(vocoder_input)

    @property
    def sample_rate(self) -> int:
        return int(self.autoencoder.sampling_rate)

    @property
    def mel_hop_length(self) -> int:
        return int(self.autoencoder.mel_hop_length)

    @property
    def mel_bins(self) -> int:
        return int(self.autoencoder.mel_bins)

    @property
    def latent_channels(self) -> int:
        return int(self.autoencoder.decoder.z_channels)

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Use a checkpoint whose audio decoder outputs 1 or 2 channels
2. Re-convert/verify the checkpoint's decoder out_ch matches the vocoder's expected layout
3. If out_ch is wrong from conversion, fix the conversion script rather than bypassing the check

ch = audio_vae.autoencoder.decoder.out_ch
assert ch in (1, 2), f"unsupported decoder out_ch {ch}"

When it happens

Trigger: Loading an audio VAE checkpoint whose decoder out_ch is neither 1 nor 2 and then decoding latents through run_vocoder.

Common situations: Experimental or mis-converted audio checkpoints, or mixing a decoder with a vocoder built for a different channel layout.

Related errors

• Vocoder is missing upsample_factor; cannot infer output samp
• Input audio must have {expected_channels} channels, got {wav
• DurationHead requires at least one of video_tokens / audio_t
• Unsupported spatial_scale {scale}. Choose from {list(mapping
• Either spatial_upsample or temporal_upsample must be True

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/aef2074715a4fb8b. Report an issue: GitHub.