MiniCPM-OCR vs Unlimited-OCR-2

MiniCPM-OCR vs Unlimited-OCR-2

Reads a document page with two OCR models and shows their results next to each other, with the detected regions linked to each text block. It suits people choosing between document OCR models.

Made by nielsr 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: This Space runs two document OCR models, MiniCPM-OCR 1.0 and Unlimited-OCR-2, on the same page and shows the results side by side. On the left you see your page with a frame around each region the app finds. On the right, each model writes its text out as blocks that link back to those regions. It runs on zero-a10g hardware at Hugging Face and is tagged MCP.

Who it suits: It suits people who need to pick an OCR model for documents, and developers who want to see how the two compare on the same page.

Industry | Who uses it | How they use it | Example
Document processing | Data and operations teams | Compare two OCR models on a real page | Upload a scanned page and check which model reads the tables better
AI engineering | ML engineers | Judge OCR output block by block | Hover a region and read the text from both models
Research | Researchers | Test OCR on papers and reports | Parse a page of a research paper and compare the Markdown

You can skip it if you only need one model's output, since the page is built for comparison.

Where to use it: You can use it in these places:

• A browser: open the Space's page on Hugging Face and nothing needs installing.
• Windows, macOS or Linux: run your own copy with Python 3.12.

Use case | Who | What they get back
Model comparison | Engineers | Two sets of text blocks for the same page
Document parsing | Operations teams | Markdown for each detected region
Output review | Researchers | Preview and Markdown source for each model

How to set it up: You need these first for a local copy:

• Python 3.12 and pip
• A GPU able to run the models, since the Space uses ZeroGPU

The online page needs none of this. For your own copy, follow the steps below.

1. Install the Hugging Face tools.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses, then the other packages.
5. Start the app.

pip install -U huggingface_hub
pip install gradio==6.31.0
pip install -r requirements.txt
python app.py

Put your own value in its key setting first.

How to use it: Open the Space in a browser, or start your own copy with the command below. Upload a document page, then read the two outputs next to the framed page.

python app.py

Tool | What it does
Page view | Shows the uploaded page with each detected region framed
Model columns | Stream each model's output as blocks linked to the regions
Preview and Markdown source | Two tabs in each column for the rendered and raw text
Copy and download | Copy a single block, or download the Markdown
GPU time and token counts | Show what each run used
/compare endpoint | Streams label, box position and Markdown for both models

Example: You upload a scanned report page. You hover over a block of text on the right, and its region lights up on the page along with the matching block from the other model.

For newcomers and experts: If you are new to this, upload a page with a table and hover over blocks in both columns. You will see at once where the two models agree and where they differ.

If you know the field, MiniCPM-OCR runs on this Space through a pipeline of orientation, layout and per-region reading. Unlimited-OCR-2 is called through its own Space, because its pinned transformers version cannot share a process with MiniCPM-OCR. Your ZeroGPU token is forwarded, so both calls use your own quota.

You upload a page
Layout finds the regions
MiniCPM-OCR reads each region
Unlimited-OCR-2 reads it elsewhere
Blocks stream in side by side

Good to know: Unlimited-OCR-2 is called through its official Space, so your page also reaches that Space. The ZeroGPU time for both models comes out of your own quota.

• License: apache-2
• Cost to run: your own ZeroGPU quota
• Data it reaches: the page you upload, which is also sent to the official Unlimited-OCR-2 Space
• Not for: private documents, since a second Space receives the page

On Hugging Face, 13 November 2026

• Likes: 8
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/nielsr/MiniCPM-OCR-vs-Unlimited-OCR-2
• Files: https://huggingface.co/spaces/nielsr/MiniCPM-OCR-vs-Unlimited-OCR-2/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.