PDF to Text

This takes the words out of a PDF as plain text. It does not run character recognition unless it has to. Most PDFs — anything exported from Word, a browser, or an accounting package — already carry their text inside them, so that text is read directly: instantly, and letter-perfect, because nothing is being guessed. Only pages that are genuinely scans get recognised, which is slower and is what the language setting is for. Everything happens inside your browser, so a contract or a bank statement never leaves your device.

How to use it

1. Add your PDF. It is read on your device — nothing is uploaded.
2. Pages that already contain text are extracted as-is, which takes about a second for a whole document.
3. Pages that are scans are recognised one at a time. Pick the language of the scan first; this only affects those pages.
4. You get the text as a file, with a marker between pages so a long document stays navigable.

Examples

• A 40-page report exported from Word: all its text out in about a second
• A scanned agreement: each page recognised in turn, with progress shown as it goes

What this tool cannot do

• Pages that already contain text are read directly and come back accurately. Scanned pages go through recognition, which is meaningfully worse — especially on poor photocopies, unusual fonts, or text at an angle.
• Recognition needs the right language selected. A page set to the wrong alphabet returns nothing rather than an error.
• Layout is not preserved. You get the words, not the arrangement of them.

Frequently asked questions

Does this work on scanned PDFs?

Yes, and it detects them for you. Each page is asked for its own text first; if there is none, the page is a scan and gets rasterised and recognised with Tesseract, running in your browser. Accuracy on a clean 300 DPI printed scan is generally above 95%, and lower on forms, faint photocopies and anything handwritten — handwriting is where free OCR is weakest and you should expect to correct it. If your PDF is a mixture, only the scanned pages take the slow path.

Can it get the images out too?

Not yet, and rather than half-do it we have left it out. Pulling the embedded pictures out of a PDF reliably means reading the image streams directly from the file, which is the work we are doing next. In the meantime our PDF to JPG tool turns each page into an image, which covers most of what people need when they ask — the difference is that you get the whole page rather than the individual photographs on it.

Why is a scanned PDF so much slower?

Because a scan is a photograph of text, and reading it means recognising every character. That runs at roughly three to six seconds a page on a laptop and several times that on a phone, so a hundred-page scan is a genuine wait — minutes, not seconds. A PDF that already has a text layer skips all of that, which is why the tool checks first rather than recognising everything by default. If your document is long and scanned, start it on a computer rather than a phone and leave the tab open.

Can it read Hindi and other Indian languages?

Yes. Hindi, Tamil, Telugu, Bengali, Punjabi, Gujarati and Marathi are all available, along with Arabic and the major European languages, and there is a combined English-and-Hindi option for the mixed documents that are common in India. Pick the language before you start, because it decides which recognition data is downloaded. Accuracy on Indian scripts is usually below what the same engine achieves on English — the letterforms are more connected — so check the result on anything that matters.

Will it give me an editable Word file?

No, and it is worth being straight about why. Getting the words out is one problem; rebuilding a document — paragraphs, headings, columns, tables, styles — is a much harder one, and even the paid converters that attempt it manage roughly 70 to 90% layout fidelity and expect you to spend twenty minutes tidying up afterwards. Rather than promise that and return scrambled columns, this gives you clean text and the images, which you can paste into Word and lay out yourself in less time than fixing a bad conversion takes.

Is my file uploaded anywhere?

No. The PDF is read by your own browser, the recognition engine runs on your device, and the zip is assembled locally. Nothing is sent to us at any point, and you can prove it: load the page, disconnect from the internet, and it still works — after the recognition data for your chosen language has downloaded once. This matters more here than on most tools, because the documents people want text out of are contracts, statements, certificates and identity papers.