PDF24 OCR: what it runs, and its options tried in OCRmyPDF
0
scripts allowed by PDF24's terms
Free, Tesseract inside, files gone in an hour
| Item | PDF24 says |
|---|---|
| Price | Free, paid for by ads, no sign-up |
| Engine | Tesseract: the page code maps Tesseract language codes |
| Languages | 123, more than one at a time |
| Options | Deskew (on), rotate, remove background, clean, force OCR, combine files |
| Output | PDF or PDF/A |
| Size | "No limits"; the page code caps a file at 1,000 MB |
| Your files | Deleted after one hour, servers in Germany |
| Desktop app | PDF24 Creator, Windows only |
| API | None that we found |
Why we did not run PDF24 itself
Its terms say: "The use of the products via automated scripts or the like is not permitted." Our tests run from scripts.
So we ran OCRmyPDF, a free tool with the same names for five of these options, on our 10-page book scan. These are OCRmyPDF's numbers, not PDF24's.
In OCRmyPDF, deskew cost 8 words
| Options | Words | Time |
|---|---|---|
| None | 1,874 | 26 s |
| Deskew | 1,866 | 41 s |
| Deskew, rotate | 1,866 | 52 s |
| Deskew, clean | 1,866 | 63 s |
| Remove background | Error |
Deskew re-saved every page as a smaller JPEG, so the file shrank from 6.5 MB to 4.7 MB. It also glued some words: ThenI, provea.
OCRmyPDF 17.13 stops on remove background: "temporarily not implemented". Our pages were straight, so deskew had nothing to fix.
PDF24 OCR not working? Four checks
- Pick the page's language. The English page starts on English, and your last choice comes back next time.
- Pages that already have text are not read again. Tick Force OCR.
- Mixed pages? Pick more than one language. In OCRmyPDF, adding German cost no words on our English pages.
- Slow? PDF24 warns that OCR is very time-consuming and points to PDF24 Creator, which is for Windows.
The same job on a Mac
brew install ocrmypdf poppler
ocrmypdf -l eng --output-type pdf scan.pdf out.pdf
pdftotext -raw out.pdf out.txt
Add --deskew only for tilted pages. For other languages,
see OCRmyPDF language.
Use -raw: plain pdftotext found 21 fewer
words, with lines out of order.
How we ran it
- A Mac with Apple M5 Max, at background priority, one OCRmyPDF run at a time with 2 pages in parallel.
- OCRmyPDF 17.13.0, Tesseract 5.5.4 with its Homebrew English model, unpaper 7.0.0 for clean.
-
The scan: 10 proofread pages from two public domain books on
Wikisource, 1,955 words. A word counts when it comes in the right
order. Text read with
pdftotext -raw. - German: the tessdata_fast model next to the English one. PDF24 does not say which Tesseract models or version its online tool uses.
- Same day, same pages: Tesseract alone 1,878 words, our Actor's code 1,879.
A PDF OCR API that allows scripts
Run it yourself. Our Apify Actor turns a scanned PDF into Markdown or text, from a script or a click. $0.006 a scanned page. Its code found 1,879 of these 1,955 words.
Or send us the scan. Tack, our AI, does the job for $29, up to 50 pages. First page free.
Questions people ask
Is PDF24 OCR free?
Yes. Its tool page says PDF24 is free of charge, paid for by ads, with no sign-up.
Which OCR engine does PDF24 use?
Tesseract, the free open source engine. The code of its OCR page maps Tesseract language codes, and its 123 languages are Tesseract models.
Is PDF24 OCR safe for private files?
PDF24 says it deletes files after one hour and that its servers are in Germany. We cannot check that from outside. A local tool keeps the file on your computer.
Does PDF24 have an OCR API?
We found none, and its terms do not allow scripts. To OCR from a script, run OCRmyPDF on your computer or call an OCR API.
Does PDF24 OCR work on a Mac?
The online tool does, in a browser. PDF24 Creator, the desktop app, installs only on Windows.