Summarising a long document in one pass does not work: a language model reads a fixed window, so feeding it a forty-page report means summarising the first two pages and calling it the whole thing. This tool splits the document instead. Paragraphs are grouped into sections of a few hundred words each, never cutting a paragraph in half, and each section is summarised on its own. What you get back is a numbered overview that follows the document's shape rather than its first page. The model itself downloads once and then runs in your browser, so the report never goes anywhere.
Features
Document split into word-bounded sections that respect paragraph boundaries
Each section summarised separately, so the whole document is covered
Short, medium and long settings for how much detail comes back
Progress shown section by section as the work runs
The model runs on your device and is cached after the first download
Headings, lists and tables all folded into the text the model reads
Numbered result you can copy or download as a text file
How to use the PDF Summarizer
Drop the PDF onto the page
Check the word count and how many sections it was split into
Choose a summary length and start
Read the numbered overview, then copy or download it
Example
Input
annual-report.pdf — 18,400 words
Output
12 numbered summaries, one per section, covering roughly the first 7,000 words.
Section boundaries follow paragraphs, so no summary starts mid-sentence.
Common errors & troubleshooting
Nothing happened for a long time on the first run. — The model downloads once, which takes a while on a slow connection. After that it is cached and later runs start immediately.
The summary stops partway through the document. — Twelve sections are summarised, which covers around seven thousand words. Split a longer document and summarise the parts separately.
A section summary repeats the opening sentence. — Small summarisation models lean on the beginning of their input. Try the longer setting, which forces more of the section into the output.
The browser ran out of memory. — Close other tabs and try again — the model needs a few hundred megabytes. A shorter document also helps.
Frequently asked questions
Where does the summarisation actually happen?
In your browser. The model is downloaded once and then runs on your device, so board papers and confidential reports are never transmitted to summarise them.
How accurate is an on-device summary?
It is a small model, so treat the result as an index rather than a substitute for reading. It is reliable at telling you what a section is about, less so at nuance.
Why summarise section by section rather than the whole file?
Because a model reads a fixed amount of text. One pass over a long report would summarise its opening pages and ignore everything else.
Can it summarise a scanned document?
Not directly — there is no text in a scan. Recognise the text first and the recognised version summarises normally.
Does the model download again every time?
No. Your browser caches it after the first run, so later sessions start straight away and work offline.
Related tools
AI Text Summarizer — Summarize long text into a short summary — the AI model runs in your browser, nothing is uploaded.
PDF Read Aloud — Listen to a PDF read out loud, passage by passage.
PDF to Text — Extract selectable text from a PDF as plain text or Markdown.
Readability Score Checker — Score text with Flesch-Kincaid, Gunning Fog, SMOG and more, with word and syllable stats.
Word & Character Counter — Count words, characters, sentences, lines, bytes and approximate tokens, live.
AI Entity & PII Extractor — Extract names, organizations, locations, emails, phone numbers and IPs from text — runs in your browser.
PDF to Word — Rebuild a PDF as an editable .docx document.