Read the video instead
A TL;DR, the key takeaways and timestamped chapters. Free, no account: the model runs in your own browser.
Recent
How it works
Captions in
The video's caption track is fetched by id, manual or auto-generated. That is the only thing the server ever handles, and it keeps a copy so the next person does not wait.
Chapters from parts
The transcript is split into roughly ten-minute parts. Each part keeps its start time, so every chapter links straight back into the video.
Folded into a summary
The chapter notes are condensed until they fit, then one last pass writes the TL;DR. The result stays in your browser, ready to copy.
FAQ
How do I summarize a YouTube video for free?
Paste the link above and press Summarize. The captions are fetched, split into parts, and a model running in your browser writes the TL;DR, the takeaways and one chapter per part. No account, no cap.
Why is it free?
The expensive part of a summarizer is the model. Unwatch runs it on your device, Chrome's Gemini Nano or a cached WebGPU model, so there is no model bill per summary and nothing to charge for.
Does it send my videos anywhere?
The summary is written on your device. The server only sees the video id it needs to fetch public captions and does not store what you summarize. Your recent summaries stay in this browser.
Which browsers work?
Chrome desktop with Gemini Nano is fastest. Chrome, Edge, Safari 26 and Firefox with WebGPU, including Safari on recent iPhones, run a small open model in the page after a one-time download.
What if the video has no captions?
Almost every video has auto-generated captions and those work fine. When there are none, open the video's Show transcript panel, copy the text and paste it into the box above.
How long a video can it handle?
Any length. Parts of about ten minutes become chapters, the notes get folded until they fit, then the summary is written. An hour-long talk takes about a minute on a recent laptop.
Is the summary accurate?
It is only as good as the captions and a small model. Auto-generated captions get names and numbers wrong sometimes, and the page says when the transcript was auto-generated. Use the timestamps to check anything that matters.
Can I share a summary?
Copy it as text or Markdown, or share the link with the video id: whoever opens it gets the same summary generated in their own browser.