AutoClip：本地把长视频自动剪成短视频的开源工具
开源项目 AutoClip 用 Whisper 本地转写、按文本打分挑片段并自动起标题，支持 YouTube 直链、封面生成和多平台发布。
Found a GitHub repo that turns a two-hour video into short clips on its own, and it does the editing on your own computer. It's called AutoClip, built by a solo developer in Shanghai who goes by zhouxiaoka, and it sits at 8,800 stars with 1,600 forks. About 600 of those stars came today. Give it a long video, a podcast, an interview, a lecture, a stream recording. It transcribes it, reads the transcript, scores every section for how good a clip it would make, then cuts the best ones out and writes a title for each. You decide how high the bar is. It works from the words rather than the pictures, which is why a normal laptop can run it. Whisper does the transcription on your machine. The scoring can run through Ollama or LM Studio, so the whole job can stay local. If you'd rather use a cloud model, it takes Gemini, Qwen, or any OpenAI-compatible key, and it sends that provider the transcript text and nothing else. A few things caught my eye. It pulls straight from a YouTube link, so there's nothing to download first. It generates a cover image for every clip. The version from yesterday can post finished clips to TikTok, Instagram, YouTube, X, LinkedIn and eight other platforms from inside the app, through a publishing service you connect. There's a desktop app for Mac and Windows, a Docker version, and a command line for people who want to automate the whole thing. The repo opens in Chinese, and the English README is one click in. The developer's GitHub bio says "I know nothing about code." The second biggest contributor to the repo is Cursor's coding agent. Four releases shipped in the last four days, so expect rough edges. A whole category of paid clipping tools charges a monthly fee for this job, and most of them want your footage on their servers to do it. Why upload your footage to someone else's server when your laptop can do this?