Anvare
All toolsAI Writing

Text Summarizer

Paste in an article, report or essay, pick how much of it to keep, and get a shorter extract built entirely from your own sentences — no rewriting, no AI model, nothing uploaded.

10% for a tight extract, 50% for a lighter trim.
0 words
—
Sentences kept—
Original word count—
Summary word count—
Estimated reading time — original—
Estimated reading time — summary—

Extractive, not AI: this is a term-frequency algorithm, not an AI or language model — it never rewrites or invents a single word. It only scores the sentences you pasted and selects the strongest ones, and it runs entirely in your browser. Nothing you paste is uploaded anywhere.

How the scoring works

Step 1 — split into sentences. The text is split on sentence-ending punctuation (. ! ?), with a short list of common abbreviations (Dr., e.g., U.S., etc.) recognised so they don't create false breaks. A line with no ending punctuation of its own — a heading, a bullet point — is treated as its own sentence too, so one-item-per-line text doesn't get fused into a single run-on sentence.

Step 2 — score each sentence. Every word is lower-cased and around 155 common English "stop words" (the, of, and, is, this, …) are ignored, since they carry no topic information on their own. Each remaining ("significant") word's frequency across the whole text is counted, and a sentence's score is the average frequency of its own significant words — so a sentence built from words that recur throughout the piece scores higher than one built from words used only once.

Step 3 — favour the opening. Scores get a small boost the earlier a sentence appears — up to 15% extra for the very first sentence, tapering to none by the last — because opening sentences conventionally carry more of a piece's topic.

Step 4 — keep the top scorers, in original order. The highest-scoring sentences are kept up to the percentage you chose, then put back in the order they appeared in your original text — so the summary reads as a compressed pass through the piece, not a shuffled list of "best bits".

FAQ

Is this an AI summarizer?
No. It's a classic extractive technique called term-frequency sentence scoring, not a language model — nothing is generated or rewritten. Every sentence in the summary is copied verbatim from your original text; the algorithm's only job is deciding which sentences to keep and what order to show them in.
Why doesn't the word-count percentage match the sentence percentage?
The slider sets what share of sentences to keep, not words — keeping, say, 25% of the sentences might keep only 22% of the words if the sentences the algorithm picked happen to be a little shorter than average (or more than 25% if they run longer). Both figures are shown after you summarize so you can see the real compression.
Is my text uploaded anywhere?
No. Splitting sentences, counting words, scoring and reordering all run as JavaScript in your browser tab. The page never sends what you paste to a server — it keeps working if you disconnect from the internet after it loads, and nothing you write is stored or transmitted.

Method

Not AI: this tool implements a well-known extractive-summarization heuristic — term-frequency sentence scoring with a small position weight — entirely in client-side JavaScript. It does not call any AI/ML model or external API, and it cannot generate new wording; it only selects and reorders sentences that already exist in the text you pasted.

Reading-time figure: the before/after reading-time estimates use an average adult silent-reading rate of 238 words per minute for English non-fiction — the pooled estimate from a meta-analysis of 190 studies and 18,573 participants. It's a population average, so any one reader will be faster or slower.