Transcribe local audio/video files offline using faster-whisper, generating plain text and timestamped outputs without OpenAI API usage. Use when the user wants transcription with zero token cost, privacy-preserving local processing, or no API key dependency.
Use this skill for local-only transcription with faster-whisper.
python3 scripts/transcribe_local.py "path/to/audio-or-video.mp4" \
--model small \
--output-dir ./output/transcribe-local
faster-whisper is installed..transcript.txt for plain text and .transcript.timed.txt for timestamps.medium or large-v3).Single file:
python3 scripts/transcribe_local.py "./input/video.mp4" \
--model small \
--language pt \
--output-dir ./output/transcribe-local
Multiple files:
python3 scripts/transcribe_local.py "./a.mp3" "./b.m4a" \
--model small \
--output-dir ./output/transcribe-local
For each input file <name>:
<name>.transcript.txt<name>.transcript.timed.txt<name>.transcript.jsonInstall Python package:
python3 -m pip install -U faster-whisper
Optional but recommended:
ffmpeg for better media format compatibility.OPENAI_API_KEY.--compute-type int8 by default for broad CPU compatibility.