Extract text

Read text out of a scan, photo or screenshot, with per-word confidence.

POST /v1/ocr/extract1 operation per image

The programmatic form of Read text in the dashboard — same models, same allowance.

Request

multipart/form-data, with your key as a bearer token.

Field	Type	Default	Description
imagerequired	file	—	JPEG, PNG or WebP.
language	string	en	ISO 639-1 language hint. Only the first two characters are used.

Example

curl -X POST https://api.quantilence.com/v1/ocr/extract \
  -H "Authorization: Bearer $QUANTILENCE_API_KEY" \
  -F "image=@photo.jpg" \
  -F "language=en"

Response

200 OK with this shape. Values below are illustrative; the keys and types are not.

{
  "success": true,
  "full_text": "INVOICE\nAcme Ltd\nTotal: $1,240.00",
  "blocks": [
    {
      "block_id": 1,
      "text": "INVOICE",
      "confidence": 0.976,
      "words": [
        {
          "text": "INVOICE",
          "confidence": 0.976,
          "bbox": [
            0.081,
            0.052,
            0.243,
            0.094
          ]
        }
      ]
    }
  ],
  "language": "en",
  "processing_time_ms": 412.7
}

Field	Type	Description
success	boolean	Always true on a 200.
full_text	string	Every block joined with newlines.
blocks	Block[]	Text grouped as the page lays it out.
blocks[].block_id	integer	Stable within one response.
blocks[].text	string	The block's words, space-joined.
blocks[].confidence	float	Mean word confidence, 0–1.
blocks[].words	Word[]	Individual words with boxes.
blocks[].words[].text	string	One word.
blocks[].words[].confidence	float	0–1.
blocks[].words[].bbox	float[4]	[x1, y1, x2, y2], normalised 0–1 against the image.
language	string	The hint that was applied.
processing_time_ms	float	Model time, excluding transfer.

Worth knowing

Boxes are normalised 0–1, so they map onto any rendered size by multiplying by width and height.
A page with no readable text returns 200 with an empty `blocks` array, not an error.