OCR Image

​

Overview

• Multi-format support - PNG, JPEG, GIF, WebP, and other common formats
• High accuracy - Advanced AI models trained on diverse document types
• Structured output - Returns both plain text and markdown-formatted results
• Layout preservation - Maintains document structure and formatting
• Handwriting recognition - Supports both printed and handwritten text
• Fast processing - Typical extraction in 2-5 seconds

• Document digitization - Convert physical documents to searchable text
• Receipt processing - Extract line items, totals, and merchant info
• Invoice automation - Pull invoice details for accounting systems
• Forms processing - Digitize handwritten or printed forms
• Business card scanning - Extract contact information
• ID verification - Read text from identity documents
• Screenshot analysis - Extract text from UI screenshots
• Data entry automation - Eliminate manual typing from documents
• Archive digitization - Convert historical documents to searchable format

​

When to Use OCR Image vs OCR PDF

• Single-page documents captured as images
• Photos of documents (receipts, business cards, signs)
• Screenshots containing text
• Scanned single pages
• Social media images with text
• Product labels and packaging

• Multi-page PDF documents
• Scanned PDFs
• Digital PDFs with embedded images
• Documents requiring page-by-page processing
• Large document sets

​

How OCR Works

1. Image Preprocessing - Image is analyzed and optimized for text extraction
2. Text Detection - AI identifies regions containing text
3. Character Recognition - Individual characters are recognized
4. Layout Analysis - Document structure and reading order are determined
5. Post-processing - Text is cleaned, formatted, and structured
6. Output Generation - Results returned in both plain text and markdown formats

​

Best Practices for OCR Quality

• Use high-resolution images (300 DPI or higher for scanned documents)
• Ensure good lighting and contrast
• Avoid blurry or out-of-focus images
• Keep text horizontal (or provide properly oriented images)

• Remove shadows and glare
• Ensure text is clearly visible
• Crop to document boundaries when possible
• Use white or light backgrounds for best contrast

• PNG or JPEG for photos of documents
• PNG for screenshots (lossless)
• JPEG for scanned documents (with high quality settings)
• Keep files under 10MB for optimal performance

​

Authentication

Authorization: Bearer ik_your_api_key_here

​

Request

​

Method 1: File Upload (multipart/form-data)

curl -X POST "https://api.incredible.one/ocr/image" \
  -H "Authorization: Bearer ik_your_api_key_here" \
  -F "file=@/path/to/receipt.png"

​

Method 2: Base64 JSON

curl -X POST "https://api.incredible.one/ocr/image" \
  -H "Authorization: Bearer ik_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "image": "iVBORw0KGgoAAAANSUhEUgAAAAUA..."
  }'

​

Request Body (JSON Method)

• image string — Base64-encoded image data (with or without data URI prefix)

​

Responses

​

Success Response

{
  "success": true,
  "text": "Extracted text from the image...",
  "method": "mistral_document_ai",
  "raw_response": {
    "pages": [
      {
        "index": 0,
        "markdown": "Extracted text in markdown format...",
        "text": "...",
        "bounding_box": {...},
        "image_url": "..."
      }
    ],
    "id": "...",
    "object": "document",
    "model": "mistral-document-ai-2505"
  }
}

​

No Text Found

{
  "success": false,
  "error": "No text could be extracted from the image",
  "text": "",
  "method": "mistral_document_ai",
  "raw_response": {
    "pages": [...]
  }
}

​

Field Reference

• success boolean — Whether text extraction succeeded.
• text string — Extracted text from the image (processed from markdown).
• method string — Always "mistral_document_ai" for OCR processing.
• raw_response object — Complete unprocessed response from Mistral Document AI, including:
  • pages array — Array of page data (single page for images)
    • index number — Page index (0 for images)
    • markdown string — Extracted text in markdown format
    • text string — Plain text version
    • Additional metadata fields
  • model string — Model used (e.g., mistral-document-ai-2505)

​

Supported Image Formats

• PNG
• JPEG/JPG
• GIF
• WebP

​

Error Responses

​

Authentication Required

{
  "error": "Authentication required",
  "message": "API key must be provided in Authorization header as 'Bearer ik_your_api_key'"
}

​

Invalid API Key

{
  "error": "Invalid API key",
  "message": "The provided API key is invalid, inactive, or expired"
}

​

Invalid Request

{
  "success": false,
  "error": "Missing 'image' or 'file' field in request body"
}

​

Understanding the Response

• Use text for simple text extraction and display
• Use markdown (in raw_response) to preserve formatting and structure
• Use raw_response when you need bounding boxes, confidence scores, or detailed metadata

​

Integration Patterns

# Extract text from uploaded documents
def process_uploaded_document(file):
    response = requests.post(
        "https://api.incredible.one/ocr/image",
        headers={"Authorization": f"Bearer {API_KEY}"},
        files={"file": file}
    )
    
    if response.json()["success"]:
        # Store extracted text in database
        text = response.json()["text"]
        save_to_database(text)
    return text

# Extract and structure receipt data
def process_receipt(receipt_image):
    result = ocr_image(receipt_image)
    
    # Use AI to structure the extracted text
    structured_data = client.answer(
        query="Extract merchant name, date, items, and total",
        response_format=receipt_schema,
        context=result["text"]
    )
    return structured_data

# Process multiple images in parallel
import concurrent.futures

def ocr_batch(image_paths):
    with concurrent.futures.ThreadPoolExecutor(max_workers=5) as executor:
        futures = [executor.submit(ocr_image, path) for path in image_paths]
        results = [f.result() for f in concurrent.futures.as_completed(futures)]
    return results

​

Performance Optimization

• Resize very large images (4000x4000+) to reduce upload time
• Use JPEG compression for photos (85-95% quality)
• Use PNG for text-heavy images and screenshots
• Crop unnecessary margins to focus on text areas

• Smaller files process faster (aim for < 2MB)
• Clear, high-contrast images process more quickly
• Parallel requests for multiple images
• Consider caching results for frequently processed images

• Batch similar documents together
• Cache results to avoid reprocessing
• Pre-process images to ensure good quality
• Use appropriate resolution (don’t oversample)

​

Error Handling

def safe_ocr_image(image_path, max_retries=3):
    for attempt in range(max_retries):
        try:
            with open(image_path, "rb") as f:
                response = requests.post(
                    "https://api.incredible.one/ocr/image",
                    headers={"Authorization": f"Bearer {API_KEY}"},
                    files={"file": f},
                    timeout=30
                )
            
            result = response.json()
            
            if result.get("success"):
                return result["text"]
            else:
                logging.warning(f"OCR failed: {result.get('error')}")
                return None
                
        except requests.exceptions.Timeout:
            logging.error(f"Timeout on attempt {attempt + 1}")
            if attempt == max_retries - 1:
                raise
        except Exception as e:
            logging.error(f"OCR error: {e}")
            raise
    
    return None

​

Common Issues and Solutions

• Solution: Check image orientation, try rotating 90/180/270 degrees
• Solution: Increase image resolution or contrast
• Solution: Verify text is actually present and legible

• Solution: Improve image quality (better lighting, focus)
• Solution: Ensure text is horizontal
• Solution: Remove shadows and glare
• Solution: Try preprocessing (contrast adjustment, noise reduction)

• Solution: Check if text is very small or low contrast
• Solution: Increase resolution
• Solution: Verify entire document is within image bounds

• Solution: Reduce image size (resize before upload)
• Solution: Compress images appropriately
• Solution: Check network connectivity

​

Performance Tips

• Use 300+ DPI for scanned documents
• Ensure good lighting and minimal glare
• Keep text horizontal and in focus
• Use high-quality camera settings

• PNG for screenshots and text-heavy images
• JPEG (85-95% quality) for photos
• Keep files under 10MB for optimal speed
• Consider preprocessing (crop, rotate, enhance contrast)

• Process multiple images concurrently
• Use threading or async for batch operations
• Implement request pooling for high-volume use

• Cache OCR results for frequently accessed images
• Store extracted text in database
• Implement deduplication to avoid reprocessing identical images