Summarize a PDF without uploading it
Zengist summarizes PDFs inside your browser. Drop the file on the page and the text is extracted on your device; on desktop Chrome, Edge and any WebGPU browser the summary is written there too, so the PDF never leaves your computer. It is free, needs no account, and takes up to 20 files in one run.
What happens to a PDF
Most PDFs carry a text layer, and that is what Zengist reads: every page, with no page limit, in a second or two. Contracts, reports, papers, statements, manuals and exported slides all work this way.
A scanned PDF is different. Its pages are pictures, so Zengist runs text recognition on each one, on your device, up to 30 pages per file. Recognition is slower than reading a text layer and less exact on faint or skewed scans, so a brief from a scan deserves a check against the original before you act on it.
Long PDFs are split on paragraph boundaries, each part is reduced to dense notes, and the notes are combined into one brief. You do not need to do anything for this; it is how a 200-page report fits a model that reads a few thousand words at a time.
How to do it
- Open zeng.ist and drop the PDF on the card, or click the paperclip and choose it. Add more PDFs if they belong together.
- Type what you want, or leave it blank for a general summary. "List every deadline and amount" gets a very different brief from "explain this in plain English".
- Press the arrow. The result shows a TL;DR, key points, every date, amount and action found, and one line per file. The label under the brief says where it ran.
Good to know
- Up to 20 PDFs per run and 50 MB per file.
- Password-protected PDFs are not read; remove the password first.
- Nothing is saved. Close the tab and the brief is gone, which is the point on a shared computer.
- On a browser that cannot run a model, the extracted text (not the file) is relayed to a cloud model and dropped. The result says so.
Tips
- For a contract or a lease, ask for obligations, deadlines and amounts by name. The model finds what you ask for.
- Pick Long in settings for dense reports; Medium is right for most documents.
- Drop the original and the amendment together and ask what changed.
Questions
Is the PDF uploaded anywhere?
Not when the run happens on your device, which is the default in desktop Chrome, Edge and any browser with WebGPU. The result label tells you. On the cloud fallback the extracted text is relayed and dropped; the file itself is never sent.
Does it work on scanned PDFs?
Yes, up to 30 pages per file, using text recognition that runs in the browser. Clean, straight scans read well; faint or crooked ones lose some words.
How long a PDF can it take?
On your device there is no page limit for a normal PDF; long files are chunked and combined. The cloud fallback accepts about 240,000 characters per run, roughly 40,000 words.
Try it now. Also: word documents, web pages, screenshots and images. Every question in one place in the FAQ; stuck on something, see help.