Vision API's Text Detection (OCR) extracts text from images, supporting both sparse text (like a street sign) via the TEXT_DETECTION feature and dense text in scanned documents via DOCUMENT_TEXT_DETECTION, which also preserves structure like pages, blocks, paragraphs, and words.
- Recognizes printed text across a wide range of languages, with automatic language detection
- DOCUMENT_TEXT_DETECTION returns structured output (pages/blocks/paragraphs/words) better suited to dense documents than the simpler TEXT_DETECTION mode
- Handwriting recognition is supported but less reliable than printed text, working best with clear, well-segmented handwriting
- For high-volume structured document processing (forms, invoices, ID documents), Google's separate Document AI product is purpose-built and generally a better fit than calling Vision API directly
Use cases
Intro for text extraction workflows using Google Cloud Vision
- Digitizing paper forms with Google Cloud Vision
- Subtitle or caption extraction from video frames via Google Cloud Vision
- Receipt data capture using Google Cloud Vision
- Invoice processing with text extraction from images
- Archival keyword indexing through text detection with Google Cloud Vision
- Compliance review by detecting sensitive terms using Google Cloud Vision
- Multilingual document analysis with Google Cloud Vision
Optional closing text