OCR PDF
The page looks the same. The difference is that search now finds things.
This tool uploads your file for conversion and deletes the temporary copy when the request finishes. What happens to your files
This tool uploads your file to a server to convert it.
Your file is uploaded to a server for this conversion. It is written to a temporary directory, used, and deleted before the response is sent — within seconds, in the same request. It is never written to a database, an object store, or a queue, and it is never logged.
What survives: Adds a searchable text layer under the existing image; the page still looks exactly the same. Accuracy depends on the scan — clean 300 DPI text is near perfect, a skewed phone photo of a receipt is not.
A quick browser check runs when you start.
How it works
- Read the live upload and retention statement.
- Choose a scanned PDF and start OCR.
- Download the searchable PDF and verify critical text.
Key benefit
Adds a searchable text layer while retaining the scanned page image.
Supported files and key limits
Input: PDF. Output: PDF.
Recognizes English only, and can be wrong on handwriting, low-resolution pages, skew, and damaged scans.
OCR accuracy starts with the scan
Clean, straight, high-resolution pages produce better results than shadowed phone photos, handwriting, or heavily compressed images. The recognition software matters too, but a poor scan limits any engine, because the detail is not in the image to begin with.
Skew, low resolution, JPEG artefacts from earlier compression, unusual typefaces and multi-column layouts all cost accuracy. Tables are a weak spot: the words are usually recognized, but the column structure is not kept, because a text layer has no cells. If you can rescan at 300 DPI, that beats any clean-up afterwards. If you cannot, proofread anything you rely on.
English text only
PDFOak’s OCR currently recognizes English. A scan in another language will usually come back with wrong characters. Use OCR software that supports your language instead.
What OCR changes, and what it skips
OCR does not redraw the page. It reads the image and writes the words into an invisible layer underneath it, so the page looks the same and search, selection and copy start to work. The file grows by a small amount, because text is tiny next to scanned images.
Pages that already contain real text are skipped, so you can run it on a mixed document without adding a second, worse layer to pages that did not need one. To check a page, try selecting a line of text. If the cursor follows the words, it already has text. If it draws a box over a picture, it does not.
Compress first, then OCR
If you also need a smaller file, compress first and run OCR last. Exact-size compression may turn a scan into page images, which removes any text layer OCR had added. Compressing first also makes the file easier to fit under the upload size limit shown above the file picker.
This one uploads your file, and takes the longest
The recognition engine is too large for a browser tab, so PDFOak uploads your file, processes it, and deletes the temporary copy when the request finishes. Every page has to be rendered and read, so a long scan takes minutes. Keep the tab open. For a confidential document, Adobe Acrobat and other desktop OCR software work on your own machine. The privacy page lists which PDFOak tools upload.