A CLI tool that transcribes Douyin (抖音) video speech to text.
- 🔗 Smart URL resolution — supports short links (
v.douyin.com), full URLs, and pasted share text - 📥 Direct CDN download — no yt-dlp needed, no cookies, no login required
- 🎙️ Dual transcription modes:
- Local mode — whisper.cpp (free, private, offline)
- Cloud mode — OpenAI Whisper API (faster, more accurate)
- 🌐 Multi-language — supports Chinese, English, and all languages whisper supports
brew install vangie/formula/douyin-transcriber- ffmpeg —
brew install ffmpeg(audio extraction) - Local mode —
brew install whisper-cpp(auto-downloads models on first run) - Cloud mode —
export OPENAI_API_KEY=sk-...
# Paste a Douyin share link — local mode (auto-detected when no API key)
dyt "https://v.douyin.com/n234HvHgzN0/"
# Full URL works too
dyt "https://www.douyin.com/video/7616660340146752819"
# Paste entire share text — URL is auto-extracted
dyt "8.46 复制打开抖音，看看【理性避难所的作品】第9集... https://v.douyin.com/n234HvHgzN0/ 05/31"
# Force local mode with language hint
dyt --local -l zh "https://v.douyin.com/abc123/"
# Cloud mode with custom endpoint (e.g., compatible API)
dyt --cloud --endpoint "https://api.example.com/v1" "https://v.douyin.com/abc123/"
# Save to file
dyt -o transcript.txt "https://v.douyin.com/abc123/"
# Keep the downloaded audio
dyt --keep-audio "https://v.douyin.com/abc123/"| Condition | Mode |
|---|---|
OPENAI_API_KEY set |
Cloud (OpenAI Whisper API) |
OPENAI_API_KEY not set |
Local (whisper.cpp) |
--local flag |
Force local |
--cloud flag |
Force cloud |
# Use a different model (tiny/base/small/medium/large-v3)
dyt --model-name small -l zh "URL"
# Use a specific model file
dyt --model-path /path/to/ggml-large-v3.bin "URL"
# Control thread count
dyt -t 8 "URL"- Resolve — short URL → follow redirects → extract video ID
- Fetch — get video info from Douyin's mobile web page (no API key/cookies needed)
- Download — grab audio from CDN directly
- Transcribe — whisper.cpp (local) or OpenAI API (cloud)
MIT