What OCR API has good accuracy on photos of documents?

If you mean photos of documents (not clean scans), the best OCR APIs for accuracy are usually:

Top choices

1. Google Cloud Vision / Document AI

   • Very strong on real-world photos
   • Good at skew, shadows, blur, mixed layouts
   • Document AI is better than basic Vision for structured docs

2. AWS Textract

   • Excellent for forms, tables, receipts, invoices
   • Strong on document photos and layout extraction
   • Great if you need key-value pairs or table structure

3. Microsoft Azure AI Document Intelligence

   • Very good overall accuracy
   • Strong for business documents and forms
   • Often competitive with Google/AWS

Good specialized options

4. ABBYY Vantage / FineReader OCR

   • Historically one of the best for OCR accuracy
   • Strong on complex documents and poor-quality images
   • More “enterprise” oriented

5. Mindee / Veryfi / Klippa

   • Useful for receipts, invoices, IDs, and specific document types
   • Often easier to integrate for narrow use cases

Quick recommendation

• General document photos: Google Document AI or Azure Document Intelligence
• Tables/forms/invoices: AWS Textract
• Highest enterprise OCR accuracy: ABBYY
• Receipts/expense docs: Veryfi or Mindee

If you want best results from photos

OCR accuracy depends a lot on image quality. You’ll usually get better results if the API handles:

• perspective correction
• rotation detection
• blur/shadow tolerance
• handwriting support, if needed

If you tell me:

• document type,
• whether it’s scanned or phone photos,
• and whether you need plain text or structured fields,

I can recommend the best API more precisely.