ToolSnap
OCRMay 15, 2026· 6 min read· Written by the ToolSnap Team

OCR Tips: How to Get the Best Text Extraction Results

OCR (Optical Character Recognition) can achieve over 99% accuracy on a clean image — but it can also produce garbled nonsense from a blurry phone photo. The difference almost always comes down to image preparation, not the OCR engine itself. Here is what you can do to get perfect results every time.

Extract text from your image — free

🔍 Try Image to Text →

What affects OCR accuracy?

OCR engines read images the way humans read — by recognizing shapes and patterns. When those shapes are clear, the engine succeeds. When they are ambiguous, blurry, or distorted, the engine guesses — and sometimes guesses wrong. The five biggest factors are:

Image resolution

Low DPI means blurry letterforms that are hard to distinguish

Lighting and contrast

Shadows, glare, and uneven light make letters hard to isolate

Text alignment

Skewed or curved text confuses line detection algorithms

Font and language

Decorative fonts and non-Latin scripts require specialized models

File format

Lossy compression (JPEG) adds noise around letter edges

Resolution: the single biggest factor

Resolution is measured in DPI (dots per inch) for scanned documents and in total pixel dimensions for photos. Most OCR engines need at least 150 DPI to function, and achieve their best results at 300 DPI or higher.

72–96 DPIScreen resolution — unacceptable for OCR. Letters look fine on screen but are too pixelated for recognition.
150 DPIMinimum viable — functional for large print, but small fonts will have errors.
300 DPIThe OCR sweet spot. Near-perfect accuracy for standard print. Recommended for all document scanning.
600 DPIBest for fine print, small fonts, or damaged documents. Files are large but accuracy peaks.

If you are using a phone camera instead of a scanner, make sure you shoot at the device's highest resolution. On most modern phones this is 12–50 megapixels — far above what OCR needs, but zoom in to fill the frame with the document rather than leaving large empty margins.

Lighting tips for phone captures

Lighting is the main differentiator between a clean capture and an unusable one when using a phone. Follow these rules:

Use indirect natural light

Place the document near a window but not in direct sunlight. Bright ambient light without harsh shadows is ideal. Overcast days produce the most even lighting.

Never use the flash

The flash creates a bright hotspot in the centre of the image and deep shadows around edges, making the document look worse than without it.

Turn on overhead lights

Overhead room lighting is often better than a phone flash. Use multiple light sources to eliminate shadows from the document edges.

Avoid shadows from your hands

Hold the phone directly above the document, not at an angle. A shadow from your hand or the phone can cover an entire line of text.

Boost contrast before uploading

Most phone galleries have a built-in contrast slider. A small boost (10–20%) on low-contrast documents significantly improves OCR results.

File format: PNG beats JPEG for text

JPEG compression is lossy — it saves space by blending nearby pixels together. This works well for photos but creates subtle artifacts around high-contrast edges like the boundary between a black letter and a white background. These artifacts confuse OCR engines.

✅ Use PNG for text documents

Lossless compression preserves sharp letter edges exactly. Perfect for screenshots, scanned documents, and any image where text legibility is critical.

⚠️ JPEG is fine for photos

If your image is a photograph with some printed text (a poster, a whiteboard), JPEG at high quality (90%+) is acceptable. Avoid low-quality JPEG settings.

Document alignment and orientation

OCR engines expect text to run horizontally in straight lines. Even a slight skew — a page tilted 2–3 degrees — reduces accuracy, and a page that is rotated 90 degrees will produce near-zero results unless the engine automatically detects orientation.

  • Keep the document flat — do not photograph from an angle
  • If using a scanner, align the page with the edge guides before scanning
  • For book pages, press the book flat against the scanner or glass
  • ToolSnap automatically corrects orientation (portrait/landscape), but straightening skew is up to you
  • Free tools like Microsoft Lens (mobile) or scanner apps auto-deskew before OCR

Frequently asked questions

Why is my OCR output full of symbols and random characters?

This usually means the image resolution is too low, or the text has very low contrast against the background. Try rescanning at 300 DPI or boosting the image contrast before uploading.

Can OCR read handwriting?

Modern OCR can handle clear, printed handwriting reasonably well — but cursive, informal, or messy handwriting remains a significant challenge. For handwriting, accuracy typically drops to 70–90% even in ideal conditions.

What languages does ToolSnap OCR support?

ToolSnap supports 50+ languages including English, French, Arabic, Spanish, German, Chinese, Japanese, and more. Language is detected automatically from the image.

My PDF is scanned — should I use Image to Text or PDF to Text?

Use Image to Text for scanned PDFs. PDF to Text only works on native PDFs that contain actual text data. If PDF to Text returns empty or garbled results, your PDF is scanned — switch to Image to Text.

Does image size affect OCR speed?

Yes. Larger images take slightly longer to process. A typical smartphone photo at full resolution takes 3–6 seconds. For bulk processing, compressing images to 2–4 megapixels before uploading speeds things up without affecting accuracy.

Try OCR on your image — free

No signup, no watermarks. Supports images and scanned PDFs.

Extract Text Free →