PDF to Text Converter

What this page extracts

Upload a PDF and the server reads its text layer, then returns the words as editable plain text. You can copy the result or download it as a .txt file named after the source document. The page also reports word and character counts as soon as the text arrives.

This is text extraction, not recognition. If a PDF contains real characters — as most exported documents, invoices, reports and ebooks do — the text comes out. If the pages are photographs of paper, there is no text layer to read, so the result will be empty or nearly so. That distinction explains most of the disappointing results people get from tools of this kind.

Where the text comes from

The file is uploaded and parsed with a PHP PDF parser. The parser walks the content streams, decodes the fonts, and reconstructs the characters according to the positioning instructions on each page. It does not render the page and it does not look at the pixels, which is why it can be fast on a long document and why it cannot help with a scan.

Because positioning information is approximate, the output is a reading-order guess rather than a layout copy. Multi-column pages may interleave, tables usually collapse into a run of values, and headers or footers can appear in the middle of a paragraph. Expect usable words, not a faithful page.

The page selection fields

Leaving every page field empty extracts the whole document. Filling any of them narrows the work, and they are applied in a fixed order of precedence:

1. Page Number — a single page. Takes priority over everything else.
2. Page Numbers — a list such as 1,3,5 or a range such as 2-6, and combinations work too.
3. Start Page and End Page — a contiguous span. Only one of the two is needed; a missing end page means the last page.
4. Page Limit — the first N pages from page one.

Selected pages are joined with a blank line between them, and the response includes the page numbers that were actually read. Out-of-range numbers are discarded silently. One quirk follows from that: if every number you supply is out of range, the selection becomes empty and the parser falls back to the first page rather than returning an error, so check the output before assuming an empty result means an empty page.

The advanced fields

Most documents need none of these. Reach for them only when the plain extraction has a specific, reproducible flaw — words fused together, columns interleaved, or a crash on a particular file.

Step by step

1. Drop a PDF on the upload area or click to browse. The file must be a PDF by type or extension.
2. Optionally fill a page field. Leave them blank for the full document.
3. Press the convert button. The status changes to Processing while the server parses the file.
4. Read the result in the output panel. Every word count and character count updates immediately.
5. Edit the text if you want, then copy it or download it as .txt.

The editor is a normal text area, so corrections stay in the download. This is useful for fixing the handful of places where the reading order slipped, without opening another program.

What comes out well

• Born-digital reports, invoices, statements and contracts.
• Ebooks and manuals with a single-column flow.
• LaTeX and word-processor exports with embedded fonts.
• PDFs generated from web pages, where the text is real characters.
• Long documents, where extracting text is far faster than rendering pages.

What comes out badly

Scanned paper is the big one. A scan is a picture; without OCR there is nothing to extract. The same applies to photographed documents and faxes. PDFs that embed text as vector outlines — some design exports and some signature blocks — also yield nothing, because the letters are drawings.

Complex layouts suffer even when the text is present: two-column academic papers, magazine pages with pull quotes, tables with merged cells, and forms where labels sit beside fields. You get the words in a plausible but not guaranteed order. Mathematical notation is often mangled, with symbols replaced or dropped, because the characters come from specialised fonts that the parser may not map.

Ligatures can appear as single characters or split, hyphenated line breaks stay hyphenated, and non-breaking spaces may survive as unusual whitespace. Clean-up is usually necessary before the text is reused.

Privacy

Unlike the image converters in this family, this tool uploads the PDF. The file is sent to the server, parsed there, and the text is returned. Do not use it for documents you would not send to a web service: identity papers, medical records, contracts under NDA or anything containing credentials. If the document must stay local, use a desktop PDF reader's own text export.

Limits

One file per request, no OCR, no batch queue, no automatic language detection, no table or column reconstruction, no layout preservation, no page images and no annotations. There is no password field, so genuinely encrypted files cannot be opened. Extraction quality depends entirely on how the PDF was produced. The output is plain text with no formatting, and the download carries no metadata.

Getting better results

• Try a single page first. It tells you quickly whether the document has a text layer.
• Use Page Numbers to skip covers and indexes. Extracting only the body avoids a lot of noise.
• Fix spacing, not content. If words run together, adjust the font space limit before editing by hand.
• Expect to clean up. Run the text through a word processor to remove repeated headers and repair hyphenation.
• For scans, use OCR elsewhere. This page will not convert pixels into words, no matter which options you change.

Common mistakes

• Assuming a blank result means a broken tool. It usually means the PDF is a scan.
• Filling several page fields at once. Only one takes effect, in the precedence order above.
• Expecting the layout to survive. Text extraction discards positioning by design.
• Uploading a confidential contract. The document leaves your device.
• Using Ignore encryption as a password cracker. It is not one.
• Trusting an out-of-range page number. The fallback can return page one instead of nothing.

Frequently Asked Questions