Marker

A local document converter turns PDFs and Office files into Markdown, JSON or HTML, with selective OCR and optional local or cloud LLM processing.

Marker is a local document converter for developers and teams preparing PDFs, scans and Office files for document processing or AI workflows. It extracts text while retaining structure such as tables, equations and page layout.

Document conversion: Convert PDFs, images, PPTX, DOCX, XLSX, HTML and EPUB into Markdown, HTML, JSON or chunks.
Structured output: Preserve tables, links, references and code blocks, represent equations as LaTeX, and save extracted images alongside the text.
Selective OCR: Read usable embedded PDF text and apply Surya OCR where needed, with balanced and fast modes that trade conversion quality against processing speed.
Optional LLM assistance: Improve table formatting, merge tables across pages and extract form values using Ollama or providers such as Gemini, Claude and OpenAI-compatible services.

Conversion runs on your hardware, using vLLM for NVIDIA GPUs or llama.cpp for CPU and Apple Silicon. The text-only mode skips OCR, so it doesn't cover scanned pages or equation recognition.

Local Ollama processing keeps the optional LLM step on your hardware; cloud providers receive the document content used in that step. Datalab's separate managed platform processes documents with Chandra rather than the local Marker pipeline. Marker's code carries the Apache 2.0 license.