The general-purpose route for mixed image formats. It keeps the source preview beside the recognized text so names, dates, totals, and identifiers can be checked in context.
Use this route when the source is an ordinary document image and the main requirement is reviewable text rather than reconstructed layout or inferred fields.
Improve the source with How to Photograph a Document for Accurate OCR, then check the next likely failure with Cropping Borders and Background Noise Before OCR. If your file or required result differs, compare PDF to Text and OCR Quality Checker.
Prepare the image before recognition
Use the highest-resolution original, not a messaging-app copy.
Crop to the document while retaining every line.
Compare consequential fields with the source before reuse.
Inspect the fields that carry meaning
| Review point | Likely failure | Validation action |
|---|---|---|
| Page shape | Perspective makes characters smaller on one side | Compare corner text and recapture square to the page |
| Small fields | Digits and punctuation lose distinct strokes | Zoom the original and verify each consequential character |
| Reading order | Labels, captions, or columns join incorrectly | Trace the intended sequence on the source before reuse |
Document-photo review example
For a photographed delivery note, crop out the desk but retain the supplier name, column headings, totals, and page edge. After recognition, compare the delivery date, reference number, quantities, and total individually. A readable paragraph does not prove that those short fields survived correctly.
When image OCR is not the right route
Use native copy and paste when the file already contains selectable text. Use an approved offline process when the source contains regulated, confidential, or high-risk information that should not be uploaded.