Audio transcription tool for multi-speaker recordings with separate audio files per speaker.
Requires: Separate audio files per speaker (tested with Craig Discord bot output)
Prerequisites: uv and ffmpeg on PATH; uv
installs the Python in .python-version. On Debian/Raspberry Pi OS:
sudo apt install ffmpeg and uv's installer. On macOS: brew install uv ffmpeg.
make check-deps verifies both.
make setup
cp .env.example .env
# Edit .env to map audio files to characters
make run folder=tmp/input/your-sessionFor testing and development, set up sample audio files:
# 1. Add sample audio files to samples/ directory
# Example: samples/jfk.wav
# 2. Create sample files in proper session structure
make create-sample-files
# 3. Test with sample environment
ENV_FILE=.env.jfk-sample make run folder=tmp/input/jfk-sample
# Individual testing steps
ENV_FILE=.env.jfk-sample make run-single file=tmp/input/jfk-sample/jfk_padded.wav
ENV_FILE=.env.jfk-sample make combine-transcriptsSee docs/README.md for complete usage instructions.
MIT. See LICENSE.