Initializing secure environment…
Initializing secure environment…
Get the key points, conclusions and main arguments out of a long document in one read. The summarisation model runs on your own device, so nothing is sent anywhere and there is no API key because there is no server. It downloads once, about 230 MB, then works offline. Choose a length — a paragraph, bullet points, or a structured brief — and every claim traces back to its page, so a summary you will act on can be checked in seconds. No sign-up, no subscription, no watermark.
To summarize a PDF free, add the file, choose a summary length, and get key points with page references. The model runs on your device, so neither the PDF nor its text is uploaded and no API key is needed. The first run downloads about 230 MB of model weights, then works offline. Free, no sign-up.
Get a structured brief first, then read the sections the brief points at. This turns a document you have been avoiding into a document you have a map of, and the map is generated from the text rather than from a marketing summary on the web.
Models paraphrase, and a paraphrase of a conditional is a different claim from the original. Numbers, dates, monetary amounts and named parties are where a fluent summary introduces an error that is invisible unless checked. Follow the page reference for anything you would act on.
Request a summary that includes open questions and unresolved points. It surfaces the gaps in a paper or a contract, which a summary of findings never does — and gaps are usually what you are looking for.
If even the extracted text may not leave your machine, this tool is not for it, and no hosted AI tool is. Extract the text and read it, or run a local model. The honest boundary is more useful than a tool that quietly works around your constraint.
No, and nothing is sent anywhere at all. The text layer is extracted in your browser and the summarisation model runs on your device, so the only thing downloaded is the model itself — about 230 MB, once, and then cached. No document text, no prompts and no questions reach any server, which is why there is no API key to supply.
No. There is no key because there is no service to authenticate against. The summarisation model — DistilBART, about 230 MB — runs on your own device through WebAssembly, so your document is summarised locally and no third party sees a word of it. The one-time cost is that model download. If you want a frontier model on a long document instead, /chat-with-pdf does that, and it tells you plainly that your extracted text goes to the provider you supply a key for.
Not until OCR has run. A scan has no text layer, so there is nothing to summarise. OCR first — and remember that the summary can only be as accurate as the recognition underneath it.
Long ones, in chunks that are then reconciled, so a 300-page report is handled. Very large documents take longer and cost more of your free-tier allowance. For a specific section, set a page range rather than the whole file.
Reliable for structure — what the document covers, its stated conclusions, its main sections. Unreliable for specifics: numbers, dates, names and conditions are where a summary quietly goes wrong. That is why page references are included. Treat a summary as a map, not as the territory.
Yes — a single paragraph, bullet points, or a structured brief covering key points, conclusions and open questions. Asking for questions it did not answer is a genuinely useful summary style, because it tells you what the document does not address.
No. The summary is in your browser, in the tab, and it is gone when you close it. This site keeps no copy of the document or the output.
The tool is free with no account here. The model call uses your own key and Google's free tier, which is generous for documents of this size. There is no paid tier of this site to upgrade to.
More ai & smart tools — all free, no upload.