faster-whisper
Local speech-to-text using faster-whisper. 4-6x faster than OpenAI Whisper with identical accuracy; GPU acceleration enables ~20x realtime transcription. SRT/VTT/TTML/CSV subtitles, speaker diarization, URL/YouTube input, batch processing with ETA, transcript search, chapter detection, per-file language map.
npx skills add majiayu000/claude-skill-registry --skill faster-whisper --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Transcribes audio locally using a faster-whisper backend, delivering transcription results with optional subtitles, speaker diarization, and multilingual support. It supports input from local files and URLs (YouTube, direct audio URLs), batch processing, and various output formats such as CSV, SRT, VTT, TTML, and transcript text. It can translate to English, auto-detect languages, and assign different languages per file via a language map. It also provides preprocessing options (denoise, normalize), stream output, and search within transcripts. It can export per-speaker audio segments when diarization is used and can detect chapters and search transcripts for terms. It emphasizes local processing without API costs and supports offline operation after model download.
How it works
The skill orchestrates local transcription using the faster-whisper backend. It exposes commands like ./scripts/transcribe <input> with optional flags to enable features such as --diarize, --format <format>, --translate, --language-map, --multilingual, --initial-prompt, --normalize, --denoise, --stream, --clip-timestamps, --export-speakers DIR, --search "term", --detect-chapters, --export-speakers DIR, and --rss URL. It supports batch processing via glob patterns, per-file language mapping via --language-map, and multiple output formats simultaneously when requested. It also integrates URL input via yt-dlp for downloading YouTube content and can export transcripts to various data formats (CSV, JSON) and subtitle formats (SRT, VTT, ASS, LRC, TTML). The toolchain ensures outputs are written to files or streamed to stdout as appropriate, with ETA and batch progress reported during batch runs. It enforces a default fast path: the basic transcription command ./scripts/transcribe audio.mp3 and adds flags only if explicitly requested by the user.
When to use it
Use this skill for local, offline transcription needs where you want speed and flexibility:
- Transcribe audio/video files and generate subtitles
- Identify speakers and diarize who spoke when
- Transcribe from YouTube URLs or podcast feeds
- Batch process multiple files with per-file language handling
- Translate to English or perform multilingual transcription
- Preprocess noisy audio, stream results, or extract chapters
- Search transcripts or export speaker-specific audio
What it can touch
The skill interfaces with the local filesystem for input/output and relies on the ./scripts/transcribe workflow, which accepts flags and file paths. It also references optional tools like ffmpeg and yt-dlp for preprocessing and URL input. It calls the faster-whisper pipeline and can produce outputs in formats including SRT, VTT, ASS, LRC, TTML, CSV, and plain text transcripts. It supports per-file language mapping via --language-map and can export per-speaker audio via --export-speakers DIR when diarization is used.
Caveats
The skill documents local execution and requires model download for offline operation. It notes that certain features (diarization, per-file language mapping, specific subtitle formats, and per-speaker exports) depend on explicit user requests and appropriate flags. It also specifies optional dependencies like ffmpeg and yt-dlp for full URL-input and processing capabilities. No outcomes are promised beyond the commands and capabilities described; behavior depends on input and flags provided.
# 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: - **Transcribe audio/video files** — meetings, interviews, podcasts, lectures, YouTube videos - **Generate subtitles** — SRT, VTT, ASS, LRC, or TTML broadcast-standard subtitles - **Identify speakers** — diarization labels who said what (`--diarize`) - **Transcribe from URLs** — YouTube links and direct audio URLs (auto-downloads via yt-dlp) - **Transcribe podcast feeds** — `--rss <feed-url>` fetches and transcribes episodes - **Batch process files** — glob patterns, directories, skip-existing support; ETA shown automatically - **Convert speech to text locally** — no API costs, works offline (after model download) - **Translate to English** — translate any language to English with `--translate` - **Do multilingual transcription** — supports 99+ languages with auto-detection - **Transcribe a batch of files in different languages** — `--language-map` assigns a different langu
- When to Use
- Quick Reference
- Model Selection
- Model Table
- Custom & Fine-tuned Models
- Load a local CTranslate2 model
- Convert a HuggingFace model to CTranslate2
- Load a model by HuggingFace repo name (auto-downloads)
- Custom model cache directory
- Setup
- Linux / macOS / WSL2
- Platform Support
- GPU Support (IMPORTANT!)
- Usage
pip install ctranslate2 ct2-transformers-converter \ Base install (creates venv, installs deps, auto-detects GPU) With speaker diarization support For CUDA 12.x uv pip install --python .venv/bin/python torch --index-url https://download.pytorch.org/whl/cu121 For CUDA 11.x uv pip install --python .venv/bin/python torch --index-url https://download.pytorch.org/whl/cu118 Basic transcription SRT subtitles
What does the faster-whisper skill do?
Local speech-to-text using faster-whisper. 4-6x faster than OpenAI Whisper with identical accuracy; GPU acceleration enables ~20x realtime transcription. SRT/VTT/TTML/CSV subtitles, speaker diarization, URL/YouTube input, batch processing with ETA, transcript search, chapter detection, per-file language map.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill faster-whisper --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.