Summary
The words out of a short video.
Two things people want from a Douyin link — the clip without the watermark, and what was actually said in it — come back from the link alone, with no download and no player in between.
What it is
An extraction tool for Douyin short videos: parse a share link into title and ID, get the watermark-free download URL, and run the audio through speech recognition to get the spoken script back as text.
What you get
- A share link parsed into the video's title and ID
- A watermark-free download link for the video itself
- The spoken script extracted from a video, from link to text in one call
- Local audio files and audio URLs transcribed on their own
- Audio too long or too large for one pass split and processed in segments
Requirements
Uvx to run it. Parsing a link and getting the download URL need no key; transcription runs on SiliconFlow's SenseVoice and needs an API key set as API_KEY, with a DashScope key accepted instead.
Setup effort
One command plus a key — uvx douyin-mcp-server, then supply credentials