Text recognition and structured extraction examples
Scenario: Extract reimbursement information from an invoice
Upload the invoice image and name the fields you need:
Recognize this invoice and return the invoice number, issue date, buyer, seller, amount, and tax as structured data.
The agent uses receipt recognition and returns structured JSON. Fields vary by document type, and missing or uncertain values are not invented.
Scenario: Extract plain text from an image
Extract all text from this meeting whiteboard photo. Return plain text only and do not save it to the cloud.
The agent runs general OCR on one image and returns the text directly. Request Markdown instead when headings, lists, or other structure should be retained.
Scenario: Extract mathematical formulas from an image
Recognize the mathematical formulas in this test and stitch the cropped formula regions into one image.
The agent detects formula regions and returns one stitched PNG. This capability is distinct from general text OCR.