LightOnOCR-3 Demo

LightOnOCR-3 Demo

Reads a page image and returns its text as Markdown, with layout grounding. A demo of the LightOnOCR-3 models that runs in the browser on ZeroGPU.

Made by lightonai on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: LightOnOCR-3 Demo turns a page image into Markdown text. It does layout grounding, so the result stays tied to how the page is laid out. It demonstrates the LightOnOCR-3 model, which the Space names in 4B, 1B and 0.8B sizes. The Space runs on Hugging Face, so you can try it in a browser with nothing to install.

Who it suits: People who need the text of scanned or photographed pages in an editable form, and developers who want to try an OCR model before building on it.

Industry | Who uses it | How they use it | Example
Document processing | Office teams | Turn a scanned page into Markdown text | Drop one page and read the Markdown
Records and archives | Archivists | Digitize printed pages into text | Try a page from the examples folder
Software | Developers | Compare the three model sizes on the same page | Run one page against each size the Space names
Research | OCR researchers | Look at how layout grounding ties text to the page | Open a page and inspect the grounded output

Anyone who only needs to read pages by eye can skip it.

Where to use it: It runs in these places.

• A web browser, on the Space's page at Hugging Face
• Windows, with Python 3.12
• macOS, with Python 3.12
• Linux, with Python 3.12

Use case | Who | What they get back
Convert a page to Markdown | Office teams | The page text as Markdown
Check the layout grounding | Researchers | Output tied to the page layout
Reuse the app code | Developers | A local copy with ocr.py and grounding.py to read

How to set it up: Using the Space online needs only a browser. A copy on your own computer needs the following.

• Python 3.12 and pip
• The huggingface_hub package, which gives you the hf command
• Disk space for the repository, which holds 38 files

1. Install the huggingface_hub package.
2. Download the Space's files into a local folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses, then install the packages in requirements.txt.
5. Start the app.

pip install -U huggingface_hub
hf download lightonai/LightOnOCR-3-Demo --repo-type space --local-dir LightOnOCR-3-Demo
cd LightOnOCR-3-Demo
pip install gradio==6.29.1
pip install -r requirements.txt
python app.py

How to use it: Open the Space's page, or run the app on your computer and open the page Gradio shows. Drop in a page image and read the Markdown that comes back.

python app.py

Tool | What it does
app.py | The main file, it builds the Gradio page
ocr.py | Holds the code that reads the page
grounding.py | Holds the layout grounding code
examples folder | Holds sample pages to try

Example: You drop in a scanned page of a printed report. You get back the page as Markdown, with the layout grounding applied to it.

For newcomers and experts: OCR means reading the text out of a picture of a page. The first small thing to try is the online Space with one of the sample pages.

The Space names three sizes of LightOnOCR-3, at 4B, 1B and 0.8B. The code is split into ocr.py, grounding.py and render.py. The hosted copy is set to run on Hugging Face's zero-a10g hardware.

You drop in a page image
The LightOnOCR-3 model reads the page
Layout grounding ties text to the page
The app shows Markdown

Good to know: The Space states no usage limits, so treat it as a demo you can read and adapt.

• License: apache-2
• Cost to run: none named
• Data it reaches: the page images you drop in, plus the Space files and model files that Hugging Face supplies
• Hardware: the Space is set to zero-a10g at Hugging Face, and what your own computer needs is not stated

On Hugging Face, 13 November 2026

• Likes: 90
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/lightonai/LightOnOCR-3-Demo
• Files: https://huggingface.co/spaces/lightonai/LightOnOCR-3-Demo/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.