memo-audio-transcribe

name	memo-audio-transcribe
description	Offline audio transcription into the MEMO repo. Use when the user asks to transcribe audio (mp3/m4a/wav), extract a transcript/文字稿, or turn a recording into transcript.md/transcript.txt/segments.jsonl in `01_Inbox/`.

MEMO Audio Transcribe

Transcribe local audio files offline (no upload) and write transcript outputs into the MEMO inbox.

Run

From the MEMO repo root:

python3 plugins/memo-audio-transcribe/scripts/run_transcribe.py /absolute/path/to/audio.mp3

Batch:

python3 plugins/memo-audio-transcribe/scripts/run_transcribe.py /path/a.mp3 /path/b.m4a

Outputs

For each input audio, the default output directory is:

01_Inbox/<YYYYMMDD_HHMMSS>_<audio-basename>/

Files:

• transcript.md (timestamped segments)
• transcript.txt (plain text)
• segments.v1.jsonl (structured segments)
• meta.v1.json (settings + provenance)

Notes

• Default model: Whisper small (CTranslate2), CPU int8.
• Dependencies and model files are cached under the user cache directory (not committed into git).