AT AI Tools

AI Text Summarizer

Summarize long English articles, reports and emails with the DistilBART AI model running privately in your browser. Short, medium or long summaries — no upload, no sign-up.

🔒 Runs entirely in your browser — nothing is uploaded

0 words

The first run downloads the DistilBART summarization model (about 285 MB) from Hugging Face. It is cached for next time, and your text never leaves this device.

Advertisement

Abstractive summaries generated on your device

This summarizer uses DistilBART CNN 6-6, a compact version of the BART sequence-to-sequence model that was fine-tuned to write news-style summaries. It is an abstractive model: instead of simply picking important sentences, it reads the whole passage and writes new, shorter sentences that capture the main points. The model runs with WebAssembly in a background Web Worker, so the page stays responsive, and your text is never sent to a server. The weights are downloaded from Hugging Face once and then kept in your browser cache for future visits.

That makes it safe for internal reports, research drafts, meeting notes and other documents you would not want to paste into an online AI service. There is no account, no daily limit and nothing is stored.

How long documents are handled

Transformer models can only read a limited number of tokens at once — about 1,024 for BART, roughly 600 to 700 English words. For longer input, the tool splits your text at sentence boundaries into sections of about 600 words and summarises each one. The section summaries are merged, duplicate sentences are removed and, for short and medium summaries, the merged text is summarised once more to produce a compact result. Long summaries keep one paragraph per section, which works well for reports with several topics.

Tips for better summaries

The model performs best on well-structured prose such as news articles, blog posts and reports. Remove navigation text, tables and reference lists before pasting. Choose short for a one- or two-sentence gist, medium for a paragraph and long when you need more detail. Like every AI summarizer, it can occasionally drop a key point or misstate a number, so check important facts against the original.

How to use

  1. Paste your textPaste an English article, report, essay or long email — at least 30 words.
  2. Choose a lengthPick a short, medium or long summary depending on how much detail you want to keep.
  3. SummarizePress Summarize. The model downloads once, then runs on your device while the progress bar fills.
  4. Use the summaryCheck the word-count reduction, then copy the summary or download it as a .txt file.

Frequently asked questions

Is my text uploaded?
No. The summary is generated inside your browser by a model running in a Web Worker. Only the model is downloaded from Hugging Face on first use; your text stays on your device.
Which model is used?
DistilBART CNN 6-6, a distilled version of Facebook's BART fine-tuned on CNN and Daily Mail news articles. It writes abstractive summaries — new sentences rather than copied ones. The quantised model is about 285 MB and is cached after the first download.
Does it work with other languages?
The model was trained on English text, so results in other languages are poor. Translate the text into English first with our AI Translator, then summarise it.
How long can the input be?
Up to about 60,000 characters. Long texts are split into sections of around 600 words, each section is summarised, and the partial summaries are merged and condensed into the final result.
Advertisement