Picture published by the creator on Hugging Face.
indra-ocr: Read scanned pages (1.25M recognizer)
Added · Updated
Made by Datalab
View profile on Hugging Face (opens in a new tab)Turn scanned PDFs and images into searchable text, Markdown, or JSON with boxes.
For developers and document teams; the card documents CPU inference.
Model details
Original model datalab-to/indra-ocr
Input Image Output Text
Ways to use it Run locally
What it does
Indra OCR reads PDFs and images and can return Markdown, plain text, or JSON with text boxes. The model card describes printed text and handwriting, multiple languages, and pages at different angles. Its recognizer uses ONNX Runtime and can run on CPU without PyTorch or a GPU. The Python package downloads model files on first use, and an optional local viewer shows recognized lines, layout blocks, reading order, and output beside the source page. The card describes the main recognizer as 1.25M parameters and the default model files as 7.3M parameters in total.
How to use it
Install indra-ocr, then provide a PDF or image to its command-line or Python interface; receive text, Markdown, or JSON with boxes, and use the local viewer to inspect page layout and reading order.
Common use cases
- Give it a scanned PDF and receive searchable Markdown for review.
- Pass an image of a form and receive recognized text with bounding boxes in JSON.
- Open a PDF in the local viewer and inspect recognized lines and reading order over each rendered page.
Can I use this?
Requirements Python 3.10 or later on supported Linux or Apple silicon macOS; install with `pip install indra-ocr`. The package uses ONNX Runtime and downloads model files on first use. The source says research and personal use and some startup use are permitted, but the model cannot be used in competition with the Datalab API.
Caveat The card names complex tables, degraded scans, vertical text order, and pages over 100 megapixels among its limitations. The model license’s commercial permissions are restricted; review the linked terms, especially if building a competing service.
Worth knowing The model card notes reduced performance on complex tables, degraded scans, and some vertical text or very large pages. Code is Apache-2.0; model weights use a modified OpenRAIL-M license that limits certain commercial uses and bars competing use with Datalab’s API.
Why it's here
Not trending this week
Last seen on the Hugging Face trending list on 26 October 2026. It stays on BuildTube so you can still find it.
- 517 people have liked it on Hugging Face.
- It was downloaded 8,056 times in the last 30 days.
Numbers checked on 8 November 2026; not refreshed since.
Behind it
View model pageBuildTube has not run or verified this project. Everything above is written from what the creator published.
Made this?
Quick questions
- What is indra-ocr?
- Indra OCR reads PDFs and images and can return Markdown, plain text, or JSON with text boxes.
- What can I use indra-ocr for?
- Give it a scanned PDF and receive searchable Markdown for review; Pass an image of a form and receive recognized text with bounding boxes in JSON; Open a PDF in the local viewer and inspect recognized lines and reading order over each rendered page
- Can I run indra-ocr on my computer?
- It runs from a terminal on your own computer. Python 3.10 or later on supported Linux or Apple silicon macOS; install with `pip install indra-ocr`. Python 3.10 or later on Linux or Apple silicon macOS; first use downloads model files.
- What do I need to use indra-ocr?
- Python 3.10 or later on supported Linux or Apple silicon macOS; the first run downloads the model files.
- What license does indra-ocr use?
- Code is Apache-2.0; model weights use a modified OpenRAIL-M license that limits certain commercial uses and bars competing use with Datalab’s API.
More from this week
- P-H3-Image-Comfy Make and edit 8-megapixel images in ComfyUI
Turn short text requests into 8-megapixel images and edit pictures on your own graphics card.
- Idol Turn one picture of an object into a 3D Gaussian splat
Idol turns one photo, render or generated image of an object into a 3D Gaussian splat file.
- agent-desktop Let AI agents read and click desktop apps
For developers building computer-use agents: agent-desktop lets an AI agent read and operate desktop apps on Mac, Windows or Linux from a command line.
Explore by topic
Post this
Download share imageFor makers
Made this? Show it’s on BuildTube — add the badge to your README or embed the card on your site.
README badge (Markdown)
[](https://buildtube.app/p/hf-datalab-to-indra-ocr/)Embed card (HTML)
<iframe src="https://buildtube.app/embed/hf-datalab-to-indra-ocr/" width="560" height="240" style="border:0;max-width:100%" loading="lazy" title="indra-ocr on BuildTube"></iframe>