1. Start with the best available source
Use the original scan or camera file rather than a screenshot of a thumbnail. If someone sent a small preview, request the original document. When a PDF already has accurate selectable text, reuse that text instead of converting it to an image and recognizing it again.
Keep an unchanged copy before adjusting contrast, rotation or cropping. The original is your reference when a later version loses faint punctuation or handwritten marks. Give working copies descriptive names so you can compare them without guessing which file was edited.
2. Make the letters readable before uploading
For a photograph, keep the page flat, hold the camera roughly parallel to it and use even light. Avoid glare over glossy paper and shadows from your hand or phone. Check a few small words before putting the document away; a retake is easier while the source is still available.
For a scan, 300 pixels per inch is a useful starting point for ordinary printed text. IMPACT recommends higher resolution for very small print. The U.S. National Archives also specifies 300 ppi for modern textual permanent records in its own digitization requirements. These are capture references, not a guarantee that every page will recognize correctly.
Changing a DPI label on an existing low-resolution file does not create new detail. Inspect character edges and punctuation, and rescan at an appropriate resolution when the original is available.
3. Straighten the view and select a coherent region
Make the text upright and remove irrelevant borders when they compete with the page. Leave a little space around the intended text: a tight crop can cut off accents, descenders or the last digit of a number.
For columns, recognize one column or a logically connected passage at a time when full-page order is unreliable. Keep table headings with their values. Separating every tiny cell may remove the context you need to reconstruct the result.
4. Compare modest image adjustments
A small contrast adjustment may help faint text, while excessive filtering may erase it. Darkening the page can also strengthen stains, paper texture or writing from the reverse side. Compare the adjusted image with the original at the same zoom level.
Try one adjustment on a difficult passage before changing everything. If an enhancement makes the image look cleaner but removes a comma or joins two letters, it has not improved the evidence for OCR.
5. Set the language and compare recognition modes
Choose the language actually used in the selected text. i2OCR provides 128 recognition choices, including script and historical variants. For a mixed-language page, separating distinct regions can make the result easier to evaluate.
Classic is a useful starting point for clear printed material. Future can help with more varied documents and handwriting. Compare the same short passage when unsure. Judge the complete transcript, including omitted words and line order, rather than a few correctly recognized sentences.
6. Review the text against the image
Check a sample from the top, middle and bottom of the selected area. Then inspect the details that carry the most meaning. A misplaced decimal point or a missing word such as not can matter more than several ordinary spelling errors.
- Names, identifiers, dates, decimal points and units match the source.
- Every intended line is present; headers and footnotes are in sensible places.
- Columns and table rows have not been combined incorrectly.
- Unclear passages are marked for review instead of silently guessed.
- A copied quotation retains the wording and punctuation you intend to cite.
7. Save the reviewed result and reuse what worked
Save the corrected transcript before translating or rewriting it. Keep the source filename or page reference with the result. If you will process similar documents later, note the language, selected region and adjustments that helped.
The online tools accept one image up to 25 MB or one PDF up to 200 MB and 750 source pages, subject to the limit shown in the tool. PDF OCR processes one selected page or area per request. Image dimensions and processing limits also apply, so a smaller, well-captured source can be more practical than a very large file.
Questions and answers
What resolution should I scan at?
Start around 300 ppi for ordinary printed text; smaller print may benefit from a higher-resolution capture. Inspect the actual letters. Resolution alone cannot fix blur, glare or damaged originals.
Should I convert every image to black and white?
No. Thresholding can remove faint text or merge strokes. Keep the original and compare a small region before committing to a black-and-white version.
Will a larger file always give better OCR?
No. File size reflects dimensions, compression and format, not just legibility. A clear, well-framed image can be more useful than a large but blurred photograph.