ChinaDocument parsing109 languages0.9B parameters

Explore PaddleOCR-VL

Parse multilingual documents and recognize text, tables, formulas, charts, and page layout with Baidu's compact vision-language model.

Model

PaddlePaddle/PaddleOCR-VL

Input

Configure your request

Document image or PDF

Upload a scanned page, photo, or PDF document.

Recognition task

Additional instruction Optional

Parameters

First PDF page

Last PDF page

Maximum output tokens

Sample output Use generation cache

Output format

Correct page orientation Unwarp photographed pages Detect page layout