Enhancement tips for faster, cleaner OCR.

1. Resize thoughtfully

Ultra-high-resolution images slow everything down. Aim for the long edge to sit between 12002000 px. LocalKit already downscales anything larger than 2000 px, but doing it yourself (e.g., before capturing a screenshot or export) keeps files lighter and prevents accidental blurring.

• Use the OS screenshot tool instead of photographing a display with another camera.
• Prefer PNG for screenshots and SVG exports; use JPEG for photos to shrink file size without sacrificing legibility.

2. Increase contrast before uploading

OCR engines love high contrast. For scans with faint text, bump the contrast and brightness slightly. LocalKit's optional “Enhance for OCR” toggle increases mid-tone contrast and exports an optimized JPEG automatically—use it whenever files are low light, taken from a phone, or above 2.5MB.

• On macOS Preview, use the Adjust Color pane; on Windows, the Photos app has similar controls.
• Flatten filters like sepia or blue-light modes before running OCR.

3. Crop away the noise

Background gradients, margins, and UI chrome can trick Tesseract into guessing. Crop the image so only the relevant text remains. For multi-column or dense layouts, run OCR on each section separately instead of the entire page.

If you only need the contents of a highlighted snippet, zoom the document and screenshot the snippet directlyless noise means cleaner recognition.

4. Pick the right language pack

Set the Language field to match the document. Combine codes with a plus sign (e.g.,eng+spa) for multilingual text. Keeping the languages concise prevents Tesseract from guessing glyphs that aren't present.

Working with math or special characters? Use the whitelist option in Advanced settings (or the API) to narrow acceptable symbols.

5. Watch out for perspective skew

When photographing paper, keep the camera parallel to the document. If that's not possible, straighten the image with your camera roll editor or a free tool like Adobe Scan or Apple Notes. An even baseline helps Tesseract segment lines correctly.

6. Batch smarter with the API

For large archives, send files through the OCR Worker API. It reuses a single TessBaseAPI instance and caches trained data, so repeated requests fly. Pair it with the enhancement toggle in the UI to maintain parity between manual and automated runs.