菜单
Transcribe audio files locally using faster-whisper (no API key needed). Use this skill when you need to transcribe voice messages, audio files, or any speech-to-text task. Supports .ogg, .mp3, .wav, .m4a and other formats. Works offline.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
基于 SOC 职业分类
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/tcapelle/skillz --skill transcribe命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Gérer les réservations périscolaires et de restauration scolaire sur l'Espace Famille / Espace Citoyens de Chambéry. Utiliser cette skill dès que l'utilisateur mentionne : réservations cantine, périscolaire, accueil matin/midi/soir, restauration scolaire, Espace Famille Chambéry, espace-citoyens.net, ou demande de gérer les inscriptions d'Alba ou Hector. Couvre la navigation sur le site, le remplissage du calendrier de réservation, l'application de périodicités, la soumission des demandes et le suivi des démarches.
Interact with Tidal music service — search catalog, manage playlists, get favorites and recommendations. Use when user asks about their Tidal music, wants to find songs, create playlists, or get recommendations.
Control WiiM audio player and play Tidal music natively. Use when user wants to play music, control playback, adjust volume, or discover WiiM devices. Supports playing Tidal songs, albums, artists, and playlists at full Hi-Res quality.
| name | transcribe |
| description | Transcribe audio files locally using faster-whisper (no API key needed). Use this skill when you need to transcribe voice messages, audio files, or any speech-to-text task. Supports .ogg, .mp3, .wav, .m4a and other formats. Works offline. |
Local speech-to-text using faster-whisper. Runs on-device, no API key required.
uv run ${CLAUDE_SKILL_DIR}/transcribe_cli.py <audio_file>
# Transcribe with auto language detection (default model: base)
transcribe_cli.py voice.ogg
# Use a larger model for better accuracy
transcribe_cli.py voice.ogg --model small
transcribe_cli.py voice.ogg --model medium
# Force a specific language
transcribe_cli.py voice.ogg --language en
transcribe_cli.py voice.ogg --language es
| Model | Size | Notes |
|---|---|---|
| tiny | ~75MB | Fastest, least accurate |
| base | ~145MB | Good balance (default) |
| small | ~460MB | Better accuracy |
| medium | ~1.5GB | High accuracy |
| large-v3 | ~3GB | Best accuracy, slowest |
Models are downloaded on first use and cached in ~/.cache/huggingface/.
Voice messages from Telegram arrive as .oga files. Convert with ffmpeg first:
ffmpeg -i message.oga /tmp/voice.mp3 -y
uv run ${CLAUDE_SKILL_DIR}/transcribe_cli.py /tmp/voice.mp3