Show HN: CPU-only fast OCR for screenshots, images, PDFs, webpages (github.com)

A new open-source tool, **textsnap**, allows users to easily convert images, screenshots, and webpages into plaintext or markdown format—all without the need for a GPU or cloud services. Utilizing a quantized PaddleOCR-VL-1.5 model, the software operates entirely on a CPU, enabling full functionality on standard laptops. Users can simply input local files, direct URLs, or clipboard images using a single command, and the tool efficiently isolates key content to perform optical character recognition (OCR). The model requires a one-time download of approximately 890 MB but allows full offline use afterward. This development is significant for the AI/ML community as it democratizes access to powerful OCR capabilities, making it easier for users without high-end hardware to utilize advanced vision-language models. Key features include automatic markdown output to preserve document structure and the option to save results as plaintext. The program's focus on local processing ensures user privacy, as images and data are not sent to the cloud. Overall, textsnap represents an accessible and efficient solution for extracting text from various media, showcasing innovation in lightweight AI applications.