Document Intelligence
Visual Document AI
Turn invoices, receipts & IDs into structured data — with bounding boxes over every field, confidence scores, and a side-by-side of OCR+regex vs a vision-language model.
model gpt-4o-miniOCR-grounded boxesconfidence scores
Annotated document
high · medium · low confidence.
Detected fields
Every document — sample or upload — runs the same real pipeline: tesseract.js OCR for boxes, gpt-4o-mini for the values. Boxes are grounded on the page by OCR.