OCR API with JSON output and bounding boxes

If you want an OCR API that returns JSON output with bounding boxes, the main options are:

Common OCR APIs

1. Google Cloud Vision OCR

• Returns text plus bounding boxes for pages, blocks, paragraphs, words, and symbols.
• Output is JSON via API response.
• Good for general-purpose OCR and document layout extraction.

2. Amazon Textract

• Strong for documents/forms/tables.
• Returns structured JSON with bounding boxes for words, lines, key-value pairs, and tables.
• Best if you need document analysis, not just plain OCR.

3. Azure AI Vision / Document Intelligence

• OCR with JSON response and bounding boxes.
• Good support for layout, forms, invoices, and multi-page documents.

4. Tesseract OCR

• Open-source engine.
• Can output bounding boxes and confidence values, but you typically wrap it yourself to produce JSON.
• Best if you want self-hosted and low cost, but less polished than cloud APIs.

5. Mindee / OCR.Space / Abbyy / Nanonets

• Various OCR APIs that provide JSON and bounding boxes.
• Often easier to integrate for specific document types.

What to look for

When choosing, check whether the API returns:

• Bounding box coordinates for each word/line/block
• Confidence scores
• Page number
• Reading order
• Structured elements like tables, key-value pairs, or paragraphs

Example JSON shape

A typical OCR response might look like:

{
  "text": "Hello world",
  "pages": [
    {
      "page": 1,
      "width": 2480,
      "height": 3508,
      "blocks": [
        {
          "text": "Hello",
          "confidence": 0.99,
          "bbox": [100, 200, 180, 240]
        },
        {
          "text": "world",
          "confidence": 0.98,
          "bbox": [190, 200, 280, 240]
        }
      ]
    }
  ]
}

Recommendation

• Need document parsing / forms / tables: Amazon Textract or Azure Document Intelligence
• Need generic OCR with boxes: Google Cloud Vision
• Need self-hosted/open-source: Tesseract + your own JSON wrapper

If you want, I can also give you:

1. a comparison table,
2. a sample request/response for one API, or
3. a Python/Node example that calls an OCR API and parses bounding boxes.