Core concepts
The objects and ideas behind the dOCR API.
Document
A document is a file you upload — a PDF, image, scanned page, or DOCX. dOCR classifies each document by source kind:
- Digital PDF — has a text layer; parsed directly.
- Scanned PDF / image — no text layer; read with a vision model (OCR).
- DOCX — converted to text.
Document type
A document type defines the fields to extract. Each field has a key
(the JSON key returned), a label, and a type (string, number, date,
boolean, array, or object).
dOCR ships with built-in types — Invoice, Receipt, Bank Statement, Pay Stub, W-2 and other tax forms, Driver's License, Passport, Utility Bill — and you can create your own. See Custom document types.
Extraction
An extraction is one run of the pipeline on one document. It has a status
(pending, running, completed, failed), the outputJson result, a
confidence score, the number of pagesProcessed, and the modelUsed.
Screenshot
A screenshot is a render of a web page (url) or HTML string (html) to an
image or PDF. It has a status (pending, rendering, ready, failed), a
format, the hosted url, and the rendered width and height. See the
Screenshots guide.
Capture
A capture is the umbrella over everything you bring into dOCR — both rendered
screenshots and uploaded documents. The unified
/captures endpoints list them
together, with a kind field (screenshot or upload) to tell them apart.
Credits & usage
dOCR meters both products with a single unit: the credit. One extracted page costs 1 credit, and one screenshot render costs 1 credit. Cached screenshots and failed renders cost nothing. Your plan includes a monthly credit allowance, and usage beyond it is billed as overage. See Billing & limits.
Processing modes
Each extraction runs in one of two modes:
- Highest Quality — the most capable model; best for complex or critical documents.
- Fastest — optimized for speed and cost.
See Processing modes.