MinerU

A local document parser converts PDFs, images and Office files to Markdown and JSON on macOS, Windows and Linux, with optional remote parsing.

MinerU is a local document parser for people preparing files for AI workflows. It handles scanned PDFs, academic papers and other documents containing OCR text, tables, figures or complex layouts.

Document formats: Parse PDFs and images alongside Word, PowerPoint, Excel, OpenDocument, EPUB and web files, including HTML and MHTML archives.
Book structure: Process EPUB books in reading order and preserve their internal source links; MHTML archives are parsed as whole documents.
Local parsing server: Serve document parsing on your own machine, with model-based processing for PDFs and images. Office and web formats use a separate local parsing path.

Documents stay local by default. Optional remote parsing sends them to a remote service, and a local failure doesn't automatically trigger an upload.

MinerU runs on macOS, Windows and Linux. CPU-only processing is supported, with the basic ONNX path requiring 2 GB RAM. More demanding processing needs additional memory; some GPU paths require 16 GB RAM and at least 8 GB VRAM. Apple Silicon and NVIDIA GPUs can accelerate model-based parsing, while the CPU-capable llama.cpp path can use Vulkan acceleration.

Supported model engines include ONNX, llama.cpp, PyTorch, vLLM, lmdeploy and MLX. Hardware and engine choice affect which parsing levels are available.