Tool 02 — live
DOCUMENT SUMMARIZER Free online · extractive, no AI model
Finds the sentences that carry the document and gives them back to you word for word. No model, no GPU, no rewriting — which means it is instant, gives the same answer every time, and cannot invent a fact that was not in your text.
YOUR DOCUMENT
→ KEY SENTENCES
Paste a document and press Summarise.
KEY TERMS
WHAT THIS IS, AND WHAT IT IS NOT
Good at
- Long prose — reports, articles, transcripts, minutes. Anything with repetition and clear topic sentences.
- Speed — milliseconds. No model load, no queue, no GPU.
- Trust — every sentence is verbatim from your document. It cannot hallucinate.
- Repeatability — identical input gives identical output, every time.
Not good at
- Reading smoothly — selected sentences can start with "It was therefore rejected." That is inherent to extraction, not a bug.
- Short or dense text — contracts, specs and poetry have nothing redundant to cut.
- Synthesis — it cannot answer "so what should we do?". It selects; it does not think.
- Rewriting — a sentence is kept whole or dropped. There is no middle.
How it works: the text is split into sentences, each is turned into a TF-IDF vector of its content words, every pair is scored by cosine similarity to form a graph, and PageRank runs over that graph. Sentences that resemble many others rank highest — a good proxy for "central to the document". Near-duplicates are dropped and the rest are put back in document order.