Baidu has launched version 3.5 of its open-source OCR tool, PaddleOCR, introducing significant new capabilities. The update includes the release of PaddleOCR.js, a browser inference SDK that allows PP-OCRv5 to run directly in browsers with WebGPU and Wasm acceleration, ensuring data privacy by keeping it within the browser. Additionally, PaddleOCR now supports one-click conversion of Word, Excel, and PPT documents into Markdown format. The update also integrates a Transformers backend, providing access to 20 primary models via Hugging Face and enabling seamless switching between PaddlePaddle's static graph, dynamic graph, and Transformers modes. Furthermore, results from the PaddleOCR-VL series, PP-StructureV3, and PP-DocTranslation can now be exported in DOCX format, enhancing document processing capabilities.