Upstage Document OCR icon

Upstage Document OCR

Extract text from document images using Upstage Document OCR. Supports JPEG, PNG, BMP, PDF, TIFF, HEIC, DOCX, PPTX, XLSX, HWP, HWPX formats.

Upstage Document OCR

AI-generated

Summary

Extract text and structured OCR data from documents and images using the Upstage Document OCR API.

Inputs

  • binaryPropertyName (required) — Name of the input item's binary property containing the file to process (e.g. 'data', 'document', 'file').
  • model (required) — Selects which OCR model to use; 'ocr' is recommended as it points to the latest stable model.
  • schema — Optional parameter to format the OCR response output into a specific vendor schema (Clova or Google) or leave as the default Upstage format.
  • returnMode (required) — Determines which data to return from the OCR response: full JSON, extracted text only, pages array, words array, or confidence score data.

Output shape

a list of JSON objects per input item where each contains the specified OCR data according to 'returnMode'

Each output item is paired with its input. Supported return modes include full JSON response, plain text extraction, arrays of pages or words, or confidence-related metadata. Errors on individual items produce error objects if 'Continue On Fail' is enabled; otherwise, execution halts.

Examples

Example 1: Extract plain text from a binary PDF file attached in the 'data' property using the recommended OCR model and default schema

binaryPropertyName='data', model='ocr', schema='', returnMode='text'

Discussion