Are my files uploaded to a server?

No. The summarization model is downloaded to your browser and runs on your own device — your text never leaves it.

What can I summarize?

Plain text (.txt), subtitles (.srt and .vtt) and anything you paste. Long transcripts are handled automatically by summarizing in sections and combining them.

Can I summarize a PDF?

Not directly — the summarizer takes .txt, .srt and .vtt files, or text you paste in. For a PDF, run it through the PDF to Markdown tool first (it also runs in your browser), then paste the text here.

Which summary styles are there?

TL;DR, key points, action items and meeting notes. Pick whichever fits before you run it.

Yes — free and unlimited. There's no account and no metering because the model runs on your device, not our servers.

Why is the first summary slower?

The first run downloads the model (about 0.9 GB), then it's cached in your browser, so later summaries start quickly. A device with WebGPU makes generation much faster.

Which model does it use?

Qwen2.5-1.5B-Instruct (Apache-2.0), compiled for WebLLM/MLC and executed on WebGPU. There is no WebAssembly fallback for it: a model this size is impractical on CPU, so a browser without WebGPU is told so rather than left grinding.

AI summarizer — turn long text into a short summary

Paste long text, or drop a transcript or subtitle file, and get back a TL;DR, key points, action items or tidy meeting notes. The AI model runs entirely on your device — nothing is uploaded, so it's free, unlimited, and private.

How to use the AI summarizer

Add your text or file

Drop a .txt, .srt or .vtt file, or paste text straight in. It never leaves your browser.
Pick a summary style

Choose TL;DR, key points, action items or meeting notes — whatever fits what you need.
Summarize & save

The model runs on your device (first run downloads it, then it's cached). Watch the summary stream in, then copy it or download .md/.txt.

Why summarize here

Your files are never uploaded

Summarization runs 100% in your browser — no server, no account, no upload. Ideal for confidential notes, calls and transcripts.

Made for transcripts

Paste a meeting or interview transcript (or drop the subtitles) and get the key points, decisions and action items — the perfect companion to the transcriber.

Free & unlimited

Because inference happens on your device, there's no per-token cost and no cap. Summarize as much as you like.

The model behind it

Powered by Qwen2.5-1.5B-Instruct, an Apache-2.0 language model, run through WebLLM. It needs WebGPU, which is why this tool is desktop-first — the whole model is loaded into your graphics card and never sees a server.

Questions

Are my files uploaded to a server?: No. The summarization model is downloaded to your browser and runs on your own device — your text never leaves it.
What can I summarize?: Plain text (.txt), subtitles (.srt and .vtt) and anything you paste. Long transcripts are handled automatically by summarizing in sections and combining them.
Can I summarize a PDF?: Not directly — the summarizer takes .txt, .srt and .vtt files, or text you paste in. For a PDF, run it through the PDF to Markdown tool first (it also runs in your browser), then paste the text here.
Which summary styles are there?: TL;DR, key points, action items and meeting notes. Pick whichever fits before you run it.
Is it really free?: Yes — free and unlimited. There's no account and no metering because the model runs on your device, not our servers.
Why is the first summary slower?: The first run downloads the model (about 0.9 GB), then it's cached in your browser, so later summaries start quickly. A device with WebGPU makes generation much faster.
Which model does it use?: Qwen2.5-1.5B-Instruct (Apache-2.0), compiled for WebLLM/MLC and executed on WebGPU. There is no WebAssembly fallback for it: a model this size is impractical on CPU, so a browser without WebGPU is told so rather than left grinding.