Comfy-Org/ComfyUI · error · ValueError

VAEEncodeAudio: input audio is None (source video may have n

Error message

VAEEncodeAudio: input audio is None (source video may have no audio track).

What it means

VAEEncodeAudio.execute raises this when the audio input object is None, which typically happens when a LoadVideo node extracted no audio track from the source video and passed the None through. The node cannot encode a latent from nothing, so it fails fast with an explanatory message.

Source

Thrown at comfy_extras/nodes_audio.py:84

class VAEEncodeAudio(IO.ComfyNode):
    @classmethod
    def define_schema(cls):
        return IO.Schema(
            node_id="VAEEncodeAudio",
            search_aliases=["audio to latent"],
            display_name="VAE Encode Audio",
            category="model/latent",
            inputs=[
                IO.Audio.Input("audio"),
                IO.Vae.Input("vae"),
            ],
            outputs=[IO.Latent.Output()],
        )

    @classmethod
    def execute(cls, vae, audio) -> IO.NodeOutput:
        if audio is None:
            raise ValueError("VAEEncodeAudio: input audio is None (source video may have no audio track).")
        sample_rate = audio["sample_rate"]
        vae_sample_rate = getattr(vae, "audio_sample_rate", 44100)
        if vae_sample_rate != sample_rate:
            waveform = torchaudio.functional.resample(audio["waveform"], sample_rate, vae_sample_rate)
        else:
            waveform = audio["waveform"]

        t = vae.encode(waveform.movedim(1, -1))
        return IO.NodeOutput({"samples": t})

    encode = execute  # TODO: remove


def vae_decode_audio(vae, samples, tile=None, overlap=None):
    latent = samples["samples"]
    if latent.is_nested:
        latent = latent.unbind()[-1]

View on GitHub (pinned to 1c6d8d45b3)

Solutions

1. Use a video file that actually contains an audio track (verify with ffprobe)
2. Insert a LoadAudio (or generated-audio) node instead of relying on the video's audio output
3. Add a conditional/is-None switch (e.g. a boolean gate node) so the encode branch is skipped when audio is None

Example fix

// before
VAEEncodeAudio(audio=video_audio_output, vae=vae)

// after
# gate the branch: only encode when the video actually has audio
Switch(audio=video_audio_output, select=1 if video_audio_output is not None else 0)

def has_audio(audio) -> bool:
    return audio is not None and 'waveform' in audio and audio['waveform'] is not None

from typing import Any, TypedDict

class Audio(TypedDict):
    waveform: Any
    sample_rate: int

def is_audio(v) -> bool:
    return isinstance(v, dict) and 'waveform' in v and 'sample_rate' in v

When it happens

Trigger: Wiring a video loader's audio output (empty because the MP4 has no audio stream) into VAEEncodeAudio's audio input. Also manually connecting an optional audio socket that received no value.

Common situations: Batch workflows that mix videos with and without audio tracks; silent screen recordings; MKV/MP4 remuxes that dropped the audio stream.

Related errors

• This Controlnet needs a VAE but none was provided, please us
• Minimum cutoff must be larger than zero.
• A cutoff above 0.5 does not make sense.
• MiniMax Music3 DAV cannot encode audio
• ERROR: VAE is invalid: None\n\nIf the VAE is from a checkpoi

AI-assisted analysis of Comfy-Org/ComfyUI@1c6d8d45b3 (2026-08-14). Data as JSON: /api/errors/85705edc083ff8b9. Report an issue: GitHub.