Save a PDF to text file from the real text layer only
PDF to text reads the real text layer and saves a .txt file. It does not recognise characters in a picture. A scan, or any PDF whose pages have no text, returns an explicit empty-text error: no text was found, the file is probably a scan, and text recognition is not available. That error is the result, not a blank success.
Loading tool…
or drop / paste
How to convert a PDF to text
- Add one PDF. If it is encrypted, type the password you already know so the text layer can be read.
- Leave Page headers on if you want a line like --- Page 3 --- between pages. Turn it off if you want the text without those markers. The toggle starts on.
- Press Extract text. The page reads each page’s text layer, groups pieces into lines, and joins those lines.
- If the file has no real text, the job stops with the empty-text error. It does not invent words from the picture.
- Download the .txt file when the read succeeds. The character count is reported with the result. The PDF stays in the list.
PDF to text options explained
Page headers default to on. The marker is three dashes, the word Page, the page number, and three dashes, on its own line. Turn the toggle off when you will paste the text into a document that should not contain those markers.
Lines are built from the text layer. Pieces on the same vertical position are joined from left to right. A multi-column page can come out in a rough order, because the extractor follows position, not a layout analysis. Tables may read across in a way that does not match the visual columns.
There is no OCR and no text recognition. A photographed page has pixels, not a text layer, unless some earlier tool embedded one. Extreme compress also removes selectability by rasterising pages. This tool will then report the empty-text error, which is accurate.
What the text file contains
You get the Unicode strings the PDF already stored, in an approximate reading order. Hyphenation at line ends is kept as stored. A heading and a body paragraph may or may not be separated the way the design intended, because the extractor uses position, not the original paragraph styles.
Page headers, when enabled, make it possible to see which page a sentence came from. They are added by this tool. They were not in the PDF. Turn them off if a later step should see only the document’s own characters.
The download is text/plain, one .txt file, not a PDF and not a Word document. This page will not rebuild formatting, fonts, or images. Extract images from PDF is the page for painted images. PDF to JPG is the page for a picture of the whole page.
Scans and empty results
The empty-text error is deliberate. A silent empty file would look like success. The message says no text was found and that the PDF is probably a scan. Text recognition is not available yet. Running the job again will not change that.
A file that mixes a cover scan with later real text can still return text, because some pages have a layer. Read the result. The header lines show which pages contributed. A page with no text contributes little between its header and the next one.
Limits, formats and output
| Accepted input | |
|---|---|
| Files per batch | 1 |
| Soft size warning | 200 MB or 1,000 pages |
| Hard stop | 500 MB total in memory |
| Output | PDF, or a ZIP when there is more than one file |
| Account | None. The tool runs in this tab. |
Your file never leaves your device
Text is read from your PDF by a script running in this tab. The work runs in a Web Worker in this tab, using WebAssembly for the heavy parts, and the result comes back as a local blob. This site does not upload the file, store it, or log its name.
To verify: load the page, turn on airplane mode, and drop the file again. It still finishes. Or open DevTools, choose Network, and watch the run: there is no upload. The architecture is described on How it works.
Frequently asked questions
Why did a scan fail?
The page has no text layer. PDF to text does not recognise characters in the picture. The error says no text was found, the file is probably a scan, and text recognition is not available. PDF to JPG can still make a picture of that scan. It will not transcribe it either.
What do the page headers look like?
A line such as --- Page 3 --- between pages, when the toggle is on. It starts on. Turn it off if you want only the document text. The marker is not proof of a heading in the PDF. It is only the page index this tool inserted.
Will columns stay in columns?
Not reliably. Pieces are grouped by vertical position and then left to right. Two columns on the same baseline can be joined into one line. A complex table may read in an order that is only roughly the visual order. Check the text before you treat it as a clean export.
Does this include text inside images?
No. Only the real text layer. A logo that is a picture of a word does not contribute those letters. Extract images from PDF can save the picture. It will not read the letters either.
Can I extract one page of text?
This page reads the document’s text layer as a whole. It does not offer a page-range field. Extract pages from PDF first if you need a shorter file, then run PDF to text on that file. Page headers will number the pages of the file you give it.
Is the text uploaded to be read?
No. The read runs in this tab. The .txt file is built in memory and downloaded from there. A password you typed is used only to open the PDF for that read. Text recognition is still not part of the job, online or off.
Last updated 2026-09-26. By Tam.