faster-whisper
Installation
Summary
Local speech-to-text 4–6x faster than OpenAI Whisper with identical accuracy; GPU acceleration enables ~20x realtime transcription.
- Supports SRT, VTT, ASS, LRC, TTML, CSV, JSON, and HTML output formats; multi-format output in one pass with
--format srt,text - Speaker diarization identifies who spoke when;
--speaker-namesmaps speakers to real names;--export-speakerssaves each speaker's audio separately - Batch processing with glob patterns, directories, and URLs (YouTube, direct links); automatic ETA, skip-existing resume, and per-file language override via
--language-map - Advanced features: transcript search with fuzzy matching, automatic chapter detection from silence gaps, filler word removal, stereo channel selection, paragraph detection, and subtitle burn-in to video
- Distilled models (distil-large-v3.5 default) deliver ~6x speedup with <1% accuracy loss; supports 99+ languages with auto-detection and translation to English
SKILL.md
Faster Whisper
Local speech-to-text using faster-whisper — a CTranslate2 reimplementation of OpenAI's Whisper that runs 4-6x faster with identical accuracy. With GPU acceleration, expect ~20x realtime transcription (a 10-minute audio file in ~30 seconds).
When to Use
Use this skill when you need to: