AI tools · English text → source-linked highlights

AI extractive text summarizer

Find key sentences in English text with a local AI model. Keep the original wording, trace each selection to its source and download the highlights.

Before you start

English text: 80–12,000 characters, 3–60 sentences, maximum 256 model tokens per sentence. Select up to 3 or 5 sentences, always fewer than the source. No rewriting, translation or fact-checking. Selections can miss important context.

First use downloads the model plus the inference runtime. Allow extra time and data. A recent desktop browser is recommended; older phones and restricted networks may fail.

Model weights: 23 MB · Xenova/all-MiniLM-L6-v2

AI processing runs on your device. When you load a model, your browser contacts jsDelivr and Hugging Face to download software and model files. Those services receive connection information such as your IP address; your selected files and text are not sent to them. Model files may stay in browser cache. Clear site data to remove the cache.

Load the model when you are ready.

How this AI tool works

MiniLM creates a semantic vector for each sentence. Code ranks sentences by their similarity to the document average while reducing repetition: 70% centrality, 30% similarity to already selected sentences. The selected original sentences are shown in source order. These weights are a product heuristic, not a validated accuracy score.

TDS · 2026-09-27

Xenova/all-MiniLM-L6-v2 · 751bff37182d3f1213fa05d7196b954e230abad9 · q8 · Apache-2.0

Model and method sources

Reproduce a real model result

Load the model, insert the example text and keep the three-sentence option. On 27 September 2026, the pinned model selected sentences 1, 3 and 6 in our Chromium check. This is one smoke test, not a quality benchmark. Other browsers may produce small numerical differences.

Source example

Our library will open a quiet study room next month. The new room will provide twenty desks and reliable Wi-Fi for local students. Visitors can reserve a desk online for up to two hours each day. The library will keep its existing opening hours. Volunteers will offer a weekly workshop on finding trustworthy research sources. The workshop is free, but visitors need to book a place. Staff will review feedback after three months before deciding whether to add more desks.

Observed selection

[1] Our library will open a quiet study room next month.
[3] Visitors can reserve a desk online for up to two hours each day.
[6] The workshop is free, but visitors need to book a place.

Public inference check

Questions about this tool

Does the AI write new sentences?

It selects existing sentences using neural sentence embeddings. It does not generate or paraphrase text. Each highlight links to its original sentence so you can inspect the context.

Does the result prove the source is correct?

No. The model estimates semantic similarity; it does not verify facts or guarantee that every important point is included. Read the source before relying on a summary.