These models perform optical character recognition, extracting text from images. They can help digitize text from scanned documents, photos, and other visual media.

Best for image to text extraction: abiruyt/text-extract-ocr

For most OCR tasks, we recommend the abiruyt/text-extract-ocr model. This versatile tool makes it simple to extract plain text from a wide variety of images.

Best for document extraction: cuuupid/marker

To get clean markdown or JSON from PDF, epub, or other document formats, use Marker. It’s a pipeline of models that supports all languages, removes headers and footers, formats equations and code blocks, and more. It can also OCR text from PDFs saved in image format.

Other utilities

Some other useful models for your text extraction pipeline:

• mickeybeurskens/latex-ocr specializes in recognizing LaTeX equations from images and converting them into usable LaTeX code
• cjwbw/docentr cleans up degraded images, removing bleed-through, artifacts and smudging
• willywongi/donut extracts structured JSON data from receipts
• pbevan1/llama-3.1-8b-ocr-correction fixes OCR errors in digitized text by reconstructing the original content using LLaMA 3.1