Load Audio

This is the plain-audio counterpart to Load Audio from Video - same idea, minus the video-demuxing step. If your source is already an audio file (a voice memo, a podcast export, a music track - anything that's already a standalone audio file rather than muxed into a video container), this is the loader you want. One field, one job: pick a file, get wav_bytes out.

It's the smallest, least interesting node in the pack, and that's the point - most of this pack's actual value is in the transcription side, not the loading side. This one exists so audio-only sources don't have to go through the video loader's demux step for no reason, or get routed through the VHS bridge node when there's no VHS involved at all. If you're building a transcript-and-caption pipeline and your source audio was already extracted or recorded separately from any video, this is the correct starting point, not a workaround.

How it works

Standard upload-and-pick pattern, the same as Load Audio from Video and ComfyUI's own native audio loaders: upload through the node, or drop a file into ComfyUI's input/ folder, then select it from the dropdown. There's no demuxing step needed since there's no video track to strip out, which makes this the simpler and slightly faster of the pack's two file loaders.

The input and output that matter

• audio - the only field. A dropdown populated from files already sitting in ComfyUI's input folder.
• wav_bytes (output, WAV_BYTES) - wire it into Whisper Transcribe for a single-file transcript, or into Audio Sink if you just want to keep a copy of exactly what got loaded.

Installing it

ComfyUI Manager: search ComfyUI-TranscriptionTools, install, restart. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/royceschultz/ComfyUI-TranscriptionTools

then restart. Nothing to download for this node specifically - the Whisper checkpoint download happens over on Load Whisper Transcription Model.

Common issues & troubleshooting

The dropdown is empty. Upload a file through the node first, or drop it into ComfyUI/input and refresh - the same fix as every ComfyUI file-picker widget shaped like this one.

A file fails to load and the pack's readme doesn't spell out supported formats. The safest fallback is a plain .wav. Whisper pipelines generally expect a clean waveform going in, and this pack's own internal type is named for exactly that.

wav_bytes won't connect to a node outside this pack. Expected, not a bug - it's TranscriptionTools' own type, not ComfyUI's native AUDIO. Stay inside the pack (Whisper Transcribe, Audio Sink), or use Convert VHS Audio to WAV bytes if you need to cross into VideoHelperSuite's ecosystem instead.