Unlimited OCR – Searchable PDF: Local OCR with Live Markdown & PDF Output

I’ve released Unlimited OCR – Searchable PDF, a local Pinokio app built around Baidu Unlimited-OCR for OCR'ing images and scanned PDF documents.

Features

• 📄 OCR for PDFs and images
• ⚡ Live OCR output while each page is being processed
• 📝 Switch between Rendered Markdown and raw Markdown
• 🖼️ Displays detected figures/images from PDF pages in the rendered OCR output
• ∑ Improved rendering for equations, arrays and Greek symbols
• 🔎 Optional Searchable Scan PDF — preserves the original document appearance and adds searchable text
• 🧩 Optional Reconstructed PDF — rebuilds the document using actual text, tables, equations and extracted images
• 💻 Supports GPU + CPU offloading for lower-VRAM GPUs
• 🔒 Runs locally on your machine

PDF generation is optional. You can use the app purely for OCR and export/copy the recognized Markdown text.

I’ve tested it successfully on Windows with an NVIDIA RTX 3050 8 GB, using CPU offloading for the parts of the model that do not fit in VRAM.

The project is open source and available through the GitHub repository linked above.

Feedback is very welcome — especially results from different GPUs, operating systems and document types.