Show HN: OCR Buddy: local browser OCR for code, formulas (LaTeX) and tables (github.com)

OCR Buddy is a new Chrome extension that offers a fully local optical character recognition (OCR) solution optimized for extracting text, code, formulas, and tables directly from the browser, all while ensuring user privacy. Unlike traditional OCR systems that often rely on large language models prone to generating incorrect or "hallucinated" text when faced with ambiguity, OCR Buddy prioritizes accuracy over fluency. It leverages a traditional OCR approach built on PaddleOCR's PP-OCRv5, ensuring that every recognized token is genuinely present in the input, thereby preventing errors in critical areas like code or numerical data. This extension is significant for the AI/ML community as it demonstrates that effective OCR can be achieved without the pitfalls of generative models, particularly in sensitive contexts. By using a combination of WebAssembly and WebGPU, OCR Buddy runs entirely within the browser, allowing for fast, efficient text extraction without sending data to external servers. Users can select specific regions to capture or perform full-page OCR with high precision, while the seamless Markdown conversion of web pages enhances content accessibility for developers and researchers, creating opportunities for new workflows with LLMs. The design choices—such as avoiding network calls for model downloads and ensuring low-confidence words are flagged—underscore a practical, user-centered approach to OCR technology.