Optical character recognition — the process that turns a scanned page of text into real, selectable, searchable characters — needs enough genuine pixel detail to reliably distinguish one letter shape from another. Too little resolution and recognition accuracy drops noticeably; beyond a certain point, more resolution stops improving accuracy at all, while continuing to add file size for no benefit.
The practical range that matters
For standard printed text at a normal size, 200–300 DPI is generally the range where OCR accuracy is reliably good, matching the same range that works well for general document legibility, covered in our guide to the best DPI for scanning documents. Below roughly 150 DPI, recognition accuracy on standard text starts to degrade meaningfully; above 300, most OCR systems see no further accuracy improvement, since the text shapes are already fully resolved at that point.
Why very small text needs more resolution than standard text
Fine print, footnotes, or any unusually small text needs proportionally more DPI to give OCR the same amount of actual pixel detail per character that standard-sized text gets at a lower setting. If a document has a mix of normal and very small text and OCR accuracy matters throughout, scanning at the higher end of the useful range (around 300 DPI) gives the small text a better chance, at the cost of a larger file.
Why scanning at 600 DPI does not improve OCR accuracy further
Once a character's shape is fully and clearly resolved in the scan — which happens well within the 200–300 DPI range for normal text — additional resolution beyond that point is not adding information that helps distinguish one letter from another; it is simply capturing finer detail within a shape that was already unambiguous. This is why doubling or tripling the DPI beyond around 300 produces a much larger file with no corresponding gain in recognition accuracy.
Other factors that matter as much as DPI, or more
- Contrast between text and background — faded text or a low-contrast background hurts OCR accuracy regardless of resolution.
- A straight, unskewed scan — text at an angle is harder for OCR to process correctly than the same text scanned straight, covered in our guide to straightening a crooked scan.
- Font and print quality of the original — a clean, standard printed font recognises far more reliably than handwriting or an unusual decorative typeface, regardless of scan resolution.
Scanning specifically for a document you plan to run OCR on
- Scan at 200–300 DPI for standard text, or toward the higher end if the document includes small print.
- Ensure even lighting and good contrast, avoiding shadows or a dim, low-contrast original if you have any control over the source.
- Keep the page straight during scanning, since a skewed capture reduces recognition accuracy independent of resolution.
When your document is already scanned at a much higher DPI than needed
If you already have a document scanned at 600 DPI or higher, there is no need to rescan it for OCR purposes — the higher resolution does not hurt accuracy, only file size, so it can be used for OCR as-is, or compressed down afterwards with the PDF Compressor if the file size itself is a separate concern.
Balancing OCR accuracy against file size in a large batch
If you are scanning a large batch of documents specifically to make them searchable, standardising on 200–300 DPI across the whole batch gives reliable OCR accuracy while keeping the total archive size manageable, rather than defaulting to the highest resolution your scanner offers out of general caution.
Testing OCR accuracy on a sample page before committing to a large batch
Before scanning a large volume of documents at a chosen DPI setting specifically for OCR purposes, it is worth running one representative page through the recognition process first and checking the accuracy of the result. This catches a poor setting or an unexpectedly low-quality original before you have invested time scanning an entire batch at a setting that turns out not to work as well as expected for your specific documents.
Frequently asked questions
Does colour versus greyscale scanning affect OCR accuracy?
Greyscale is generally sufficient and sometimes preferable for OCR on plain text documents, since colour data adds nothing useful for recognising black text on a white background.
Will OCR work reliably on handwriting?
Handwriting recognition is considerably less reliable than printed text recognition regardless of scan resolution, since handwriting varies far more in shape between individuals.
Does a higher DPI ever hurt OCR accuracy?
Not directly, though an unnecessarily large file from excessive DPI can slow down processing in some OCR systems without improving the actual result.
What DPI should I use for a document I am not sure will need OCR later?
200–300 DPI is a safe general default that works well both for OCR, if it turns out to be needed, and for normal on-screen legibility.
Does the scanning device matter, separate from the DPI setting?
Yes, a scanner or camera that produces a sharp, well-focused, evenly lit capture at a given DPI will outperform a blurry or unevenly lit capture at the same nominal DPI.
Is there a minimum DPI below which OCR essentially will not work?
Below roughly 150 DPI, accuracy on standard text drops meaningfully enough that it is generally not worth relying on for anything important.
Is the tool free and private?
Yes. No sign-up, no watermark, and files are removed from the server automatically about an hour after processing.