Granite Vision Document Intelligence

Granite Vision Document Intelligence

Pulls summaries, CSV data, code, tables and image descriptions out of PDFs and images, using IBM Granite Vision and Docling in a Gradio demo.

Made by ibm-granite 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: Granite Vision Document Intelligence is a Gradio demo from IBM Granite that reads documents. You give it a PDF or an image, and it can produce a summary, CSV data, code, tables or a description of an image. It combines the Granite Vision model with Docling, and it runs on GPU hardware at Hugging Face.

Who it suits: People who work with PDFs and scanned pages and want the contents turned into usable text and data.

Industry | Who uses it | How they use it | Example
Finance | Analysts | Turn a report page into data | Extract a table from a PDF as CSV
Operations | Office teams | Summarize long documents | Upload a PDF and read a short summary
Software | Developers | Study a document pipeline built on Granite Vision | Read the files in src
Research | Document AI researchers | Try a vision model on mixed pages | Upload a page with images and tables

People who need to process large batches of files automatically can skip it, since this is a demo.

Where to use it:

• A browser, by visiting the Space on Hugging Face
• A computer of your own that runs Windows, macOS or Linux and has Python 3.12
• A local Gradio page, once src/app.py is running

Use case | Who | What they get back
Summarize a document | Office teams | A short summary of the PDF or image
Extract tables | Analysts | The table contents, with CSV data
Describe a picture | Anyone | A written description of the image
Pull out code | Developers | Code that appears in the document

How to set it up:

• Python 3.12 and pip on Windows, macOS or Linux
• The Python packages in requirements.txt
• Gradio 6.17.3, the version the Space uses
• A GPU, since the Space is set to zero-a10g

1. Install the Hugging Face hub package.
2. Download the Space with the hf command.
3. Move into the new folder.
4. Install gradio at the version the Space uses, then the listed packages.
5. Start the app from the src folder.

pip install -U huggingface_hub
hf download ibm-granite/granite-vision-document-intelligence --repo-type space --local-dir granite-vision-document-intelligence
cd granite-vision-document-intelligence
pip install gradio==6.17.3
pip install -r requirements.txt
python src/app.py

How to use it: Online, visit the Space on Hugging Face, upload a PDF or an image, and choose what to pull out. On your own computer, run the start command and open the local page Gradio serves.

python src/app.py

Tool | What it does
src/app.py | Starts the Gradio interface
requirements.txt | Lists the Python packages to install
build | A folder of build files kept in the repository

Example: You upload a PDF that contains a table and ask the demo to extract the CSV data. It reads the page and returns the table contents as data you can copy.

For newcomers and experts: If you are new, upload one short PDF and ask for a summary first. That is the quickest way to see what the demo reads and how it answers.

If you know the field, the demo pairs the ibm-granite/granite-vision-4.1-4b vision model with Docling for document handling. The trade-off is a GPU requirement, and results depend on how clean the pages are.

You upload a PDF or an image
Docling reads the document
Granite Vision interprets the content
You get a summary, table or description

Good to know:

• Hardware: the Space is set to zero-a10g
• Models: it names two Granite vision models on Hugging Face
• License: apache-2
• Cost to run: GPU hardware, no API key or paid service is named
• Data it reaches: the PDFs and images you upload, and the models it fetches from Hugging Face

On Hugging Face, 13 November 2026

• Likes: 16
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/ibm-granite/granite-vision-document-intelligence
• Files: https://huggingface.co/spaces/ibm-granite/granite-vision-document-intelligence/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.