OPF Image Anonymizer

OPF Image Anonymizer

Covers personal details in screenshots and other images with black bars you can edit, using OCR and OpenAI's Privacy Filter, for anyone who shares chats, emails or documents.

Made by ysharma 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 app hides personal information in images. It reads the text in a screenshot with OCR, runs the words through OpenAI's Privacy Filter, and draws black bars over whatever the filter marks as personal. You can then adjust the bars and save the picture.

Who it suits: Anyone who shares screenshots of chats, emails or documents and needs names, contact details or account numbers covered first. It also fits developers who want to call the detection from a script.

Industry | Who uses it | How they use it | Example
Customer support | Support agents | Cover names and emails in a chat screenshot before sharing it | A ticket image safe to pass to another team
Social media | Writers and community managers | Hide handles and contact details in a screenshot before posting | A PNG ready to publish
Legal and compliance | Analysts | Blank out account numbers and addresses in a document image | A redacted copy for review
Education | Teachers | Remove student names from a shared screenshot | Course material without personal details

If your images never hold personal details, you can skip it.

Where to use it:

• A web browser: open the Space on Hugging Face
• Windows, macOS or Linux with Python, for your own copy
• A Python script, through the Gradio API call the README shows
• The clipboard: paste an image in and copy the result out

Use case | Who | What they get back
Redact a screenshot | Anyone sharing chats or emails | The image with black bars over detected personal text
Tune the bars | Careful editors | Bars moved, added or deleted until the picture is right
Detect in a script | Developers | Image size and a list of boxes with labels and text

How to set it up: You need Python and pip for your own copy. The README names a GPU for the filter, and Hugging Face hosts the Space on zero-a10g hardware.

• Python with pip, on Windows, macOS or Linux
• Gradio 6.13.0, the version this project pins
• The Python packages listed in requirements.txt
• The system packages listed in packages.txt, which the Space installs with apt-get. The README names Tesseract for OCR

1. Install the Hugging Face hub package.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio 6.13.0 and the packages in requirements.txt.
5. Run app_v2.py.

pip install -U huggingface_hub
hf download ysharma/OPF-Image-Anonymizer --repo-type space --local-dir OPF-Image-Anonymizer
cd OPF-Image-Anonymizer
pip install gradio==6.13.0
pip install -r requirements.txt
python app_v2.py

How to use it: Drop an image onto the page or paste it from the clipboard. Choose which of the eight categories to hide, then check the bars. Drag to draw one, click to select it, drag to move it and press Delete to remove it. Save a PNG or copy the result.

Tool | What it does
Canvas editor | Draw, select, move and delete black bars over the image
Category toggles | Switch each of the eight personal data types on or off from the sidebar
anonymize_screenshot | The Gradio API call that returns the image size and a list of boxes

To call it from Python, the README shows this client code.

from gradio_client import Client, handle_file
client = Client("YOUR_SPACE_ID")
result = client.predict(handle_file("screenshot.png"), api_name="/anonymize_screenshot")

Example: Say you paste a screenshot of a support chat. The app finds a customer name and an email address and covers each with a black bar. You delete one bar that hid a harmless word, then save the PNG.

For newcomers and experts: Personal information means things like a name, an address or an account number. To try the app, open the Space, drop in any screenshot and look at where the black bars land. Nothing needs installing online.

Under the hood, Tesseract returns each word with its box, and the app rebuilds the full text while remembering which characters belong to which box. The filter reads that text in a single forward pass with a 128k context. Detected character spans are matched back to words, grouped by line and merged into padded rectangles. Because boxes come from OCR, words it misses may stay uncovered, so review the result.

You drop or paste an image
OCR finds each word and its box
The privacy filter marks personal text
Marked words become black bars
You edit the bars, then save or copy

Good to know:

• License: apache-2, as listed on Hugging Face
• Cost to run: none named. The hosted Space uses Hugging Face's zero-a10g hardware
• Data it reaches: your image goes to the backend for detection. Editing and export stay local, and you can paste from or copy to the clipboard
• Not for: a final guarantee of redaction. You can edit the bars, so review them before sharing

On Hugging Face, 13 November 2026

• Likes: 20
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/ysharma/OPF-Image-Anonymizer
• Files: https://huggingface.co/spaces/ysharma/OPF-Image-Anonymizer/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.