Nodes/ComfyUI-TranscriptionTools/Load Audio from Batch
ComfyUI Node

Load Audio from Batch

Transcribe a whole folder of files in one pass

By royceschultz·Created 2 years ago·Updated about a year ago· 35
Load Audio from Batch
    • wav_bytes_batch
    • count
    • filenames
    base_folder
    path
    regex_match
    regex_exclude

    This is the pack's batch entry point - the node that makes good on the readme's "batch processing of many files at once" claim. Instead of loading one audio or video file per node the way Load Audio or Load Audio from Video do, you point this at a folder, optionally filter it with a couple of regexes, and it hands you every matching file back as one bundle, ready for Whisper Transcribe Batch.

    How it works

    It scans a directory, checks each filename against your patterns, and reads every match into one WAV_BYTES_BATCH payload - a single object carrying however many files it found, rather than one wire per file.

    The inputs and outputs that matter

    • base_folder - input or root. input scans relative to ComfyUI's own input/ directory. root points somewhere else on the machine entirely - useful if your source files live outside ComfyUI's normal folders.
    • path - blank by default. A subfolder inside whichever base you picked.
    • regex_match / regex_exclude - two plain regex filters, both blank by default (meaning: grab everything). Use regex_match to narrow it down - say, to just .wav files if video clips are sitting in the same folder - and regex_exclude to drop a subset by name.

    Three outputs: wav_bytes_batch (feeds straight into Whisper Transcribe Batch), count (INT, how many files it actually found - a quick sanity check before you kick off a long run), and filenames (LIST[STRING], the matched filenames in order - your way of correlating results back to source files, since Whisper Transcribe Batch itself doesn't hand back per-file text sockets).

    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. No separate model download for this node - that happens over on Load Whisper Transcription Model.

    Common issues & troubleshooting

    count comes back 0. Almost always the base_folder/path combination pointing somewhere that doesn't actually have your files, or a regex_match that's stricter than you meant. Test with both regex fields blank first to confirm you're pointed at the right folder, then add filters back one at a time.

    root doesn't find files you know are there. Remember it's not scanning ComfyUI's input/ directory by default - if your files are actually sitting in ComfyUI/input, use the input option instead, or give root the correct absolute path.

    On a hosted executor, root is a trap if you're not sure what it resolves to. The filesystem a run sees on a cloud executor isn't your laptop - favor input (files you actually uploaded into the workflow) unless you know exactly what root points at on that machine.

    Categoryaudio

    Inputs (4)

    NameTypeDefaultDescription
    base_folderCOMBO2 options: input, root
    pathSTRING
    regex_matchSTRING
    regex_excludeSTRING

    Outputs (3)

    NameTypeDescription
    wav_bytes_batchWAV_BYTES_BATCH
    countINT
    filenamesLIST[STRING]