Quick answer: To OCR a faded document, scan it in grayscale at 400–600 DPI, even out the background and boost local contrast, set the correct document language, run OCR, and save the result as a searchable PDF/A. Then check names and numbers by eye.
In this guide
- Why faded documents break OCR
- Scan it right
- Restore the image before OCR
- Run OCR and get a searchable PDF
- Know the limits
- Save as PDF/A
- Organize your archive
- Checklist
- FAQ
Why faded documents break OCR
OCR was built for clean, modern print. Old documents break its assumptions: pale ink, yellowed or stained paper, unusual typefaces and spelling, multiple columns, stamps, margin notes and handwriting.
A page that looks readable to your eyes can still produce gibberish. In most cases the problem is the image, not the OCR engine, so preparation matters more than any setting.
Scan it right
- Resolution: 400–600 DPI. Faint ink needs the detail.
- Mode: Grayscale or color. Pure black-and-white scanning throws away the faint strokes you're trying to save.
- Format: TIFF or PNG for master files, not heavily compressed JPEG.
- Position: Press the page flat and avoid shadows near the spine.
- Copies: Keep the originals untouched and work only on copies.
Restore the image before OCR
The goal is more contrast between ink and paper without damaging the letter shapes:
- Local contrast: Enhance region by region instead of brightening the whole page, so a dark stain doesn't wash out faint text elsewhere.
- Flatten the background: Yellowed paper has uneven tone. Evening it out makes faded ink stand out across the whole page.
- Adaptive thresholding: One global black/white threshold fails on aged paper. Adaptive thresholding decides pixel by pixel based on the neighborhood.
- Don't overdo it: Heavy sharpening breaks thin strokes and merges neighboring letters. Test on your five worst pages first. Sometimes a cleaned grayscale image beats hard black-and-white.
Run OCR and get a searchable PDF
- Open the free OCR tool.
- Upload your cleaned scan or PDF.
- Choose the document language. This is the most important setting. Old German text recognized as English fills with errors.
- Turn on deskew and auto-rotate if your pages are tilted.
- Run OCR and download the result.
What you get: The original image stays as the visible layer, and an invisible text layer sits on top, with each word positioned where it appears. The page looks identical to the scan, but you can search, highlight and copy text. Stamps, stains and handwriting are preserved.
Know the limits
OCR is excellent on clean print, decent on regular print, and unreliable on the rest.
| Content | What to expect | What to do |
|---|---|---|
| Clean printed text | Very good | Spot-check |
| Multi-column newspapers | Columns may interleave | Split into column images, OCR separately |
| Stamps, margin notes | Often skipped or mixed in | Review by eye |
| Handwriting | Rough draft at best | Use specialized HTR tools and verify manually |
| Names, dates, numbers | Highest risk | Always check (1/7, rn/m, O/0) |
Save as PDF/A
PDF/A is the archival version of PDF. It embeds fonts and forbids features that might stop working in the future. After OCR, convert your final files with the PDF to PDF/A converter.
A solid setup:
- Masters: Original TIFF/PNG scans, untouched
- Access copy: Searchable PDF/A
- Text copy: Plain text of the OCR output for search and backups
Organize your archive
- Consistent names:
collection_year_type_page, e.g.,parish-register_1912_baptisms_042 - One folder per source, with originals, access copies and text in separate subfolders
- OS search: Modern systems index text inside PDFs, so a surname search finds every page
- Simple index spreadsheet: Date, source, description, OCR status (checked / needs review)
- Three backups: Two local, one off-site
Checklist
- Scanned at 400–600 DPI in grayscale or color
- Originals saved and untouched
- Background evened, contrast improved locally
- Correct language selected
- OCR run, names and dates checked by eye
- Final files converted to PDF/A
- Three backups created
FAQ
Can OCR read very faded text?
Often yes, if you improve contrast first. Extremely faded ink may need several enhancement attempts.
What DPI should I use for old documents?
400–600 DPI works for most historical material.
Can OCR read handwriting?
Standard OCR works best on print. Handwriting needs specialized tools and human review.
What is a searchable PDF?
A PDF that shows the original scan but has a hidden text layer, so you can search and copy text.
What is PDF/A?
An archival PDF format designed to stay readable for the long term.
Is it safe to upload sensitive documents to online OCR tools?
It depends on the service. Check how long files are stored, how they're protected, and what the terms say about reuse. When in doubt, test with non-sensitive pages first.
Try it on your own documents
Pick ten pages, clean them up, run them through the OCR tool, and see what works. Acrion has 50+ other free PDF tools if you need to merge, compress or convert along the way.