📄 PaddleOCR-VL Demo

PaddleOCR-VL-0.9B — multilingual (109 languages) document parsing via a compact vision-language model. Upload an image and pick a task. Element-level recognition: text, tables, formulas, charts.

Model: PaddlePaddle/PaddleOCR-VL · License: Apache 2.0

💡 Element-level recognition (this demo) works best on a single cropped element. For full-page document parsing with layout analysis, use the official PaddleOCR pipeline.