transcribe

Local transcription using Whisper.

Usage

• pixi run triton transcribe local <audio-or-dir> --output-dir outputs/transcripts --model tiny

Options

• --model: whisper model size (tiny, base, small, medium, large)
• --device: auto|cpu|cuda
• --compute-type: auto|float16|float32
• --language: language code (e.g., en)
• --write-json: write segments to JSON