douyin-video

douyin-video downloads watermark-free Douyin videos and extracts their spoken transcripts. It solves the problem of turning a Douyin share link into usable assets: it parses the link to find the direct, watermark-free video URL (no API key required), downloads the video locally, and uses speech recognition to pull the spoken content out of the video and save it to a file automatically. It is intended for getting video information, downloading watermark-free videos, and batch-extracting video transcripts.

The skill is driven by a bundled Python script, `douyin_downloader.py`, with three main actions: `info` to resolve video ID, title, and download link from a share link; `download` to save the watermark-free video to a chosen directory; and `extract` to transcribe the audio and save it. Getting information and downloading need no API key, while transcript extraction uses the SiliconFlow (硅基流动) SenseVoice API and requires an `API_KEY` environment variable, plus FFmpeg installed on the system for audio/video processing. The extraction pipeline downloads the video, uses FFmpeg to pull an MP3 audio track, calls the speech-recognition API, and cleans up temporary files, saving each video's transcript as a Markdown file inside a folder named after the video ID (optionally keeping the source video with `--save-video`). The functions can also be called directly from Python via `get_video_info`, `download_video`, and `extract_text`, and a quiet mode reduces output.

The skill targets content researchers, editors, and analysts who need clean copies of Douyin videos and searchable transcripts. It is most useful for repurposing or analyzing short-video content at scale. The documentation notes it is for learning and research use only and that users must comply with applicable laws and avoid any copyright-infringing or illegal use.