Local YouTube video summarizer.
Paste a URL, get an HTML page with keyframe screenshots and AI summaries. Runs entirely on your machine.
Photo by Jakob Owens on Unsplash
URL → yt-dlp → Whisper → PySceneDetect → FFmpeg → Ollama → HTML
Downloads the video, transcribes it, detects scene changes, pulls a keyframe at each cut, summarizes each section through a local LLM, and renders a self-contained HTML file with everything embedded. Also includes a shorts creator for vertical clips with word-level captions, and a GUI if you don't want the CLI.
Requires Python 3.10+, yt-dlp, FFmpeg, and Ollama.
ollama pull mistral
setup.bat
run.bat https://www.youtube.com/watch?v=VIDEO_ID --openpython yt2.py URL [OPTIONS]
--model TEXT Ollama model (default: mistral)
--whisper-model TEXT tiny.en / base.en / small.en / medium.en / large-v3
--scene-threshold FLOAT Scene sensitivity (default: 27.0)
--no-cache Force re-process
--open Open result in browser
MIT