The Glaze Project
What does it do?
- AI Art Protection
- Style Preservation
- Digital Rights Management
- Creative Content Protection
- Adversarial Machine Learning
How is it used?
- 1. Download & install
- 2. Input image/s
- 3. Adjust settings
- 4. Preview changes
- 5. Run Glaze
Who is it good for?
- Graphic Designers
- Creative Content Creators
- Visual Artists
- Digital Illustrators
- Fine Art Photographers
Details & Features
-
Made By
University of Chicago -
Released On
Glaze is a software tool designed to protect artists' unique styles from being replicated by artificial intelligence models. It applies subtle modifications to artworks that are imperceptible to human viewers but effectively disrupt AI training processes, preventing accurate replication of an artist's distinctive features.
Key features:
- Style Cloaking: Applies subtle changes to artworks that confuse AI models without being noticeable to human eyes.
- Intensity Adjustment: Allows users to specify the strength of the style cloak, balancing protection and visual impact.
- Render Quality Selection: Offers options for render quality, affecting processing time and protection effectiveness.
- Preview Functionality: Enables artists to view the Glazed result before finalizing changes.
- Error Handling: Provides feedback and suggestions if protection is insufficient.
- Cross-Platform Support: Available for Windows and MacOS with GPU acceleration.
- WebGlaze: A web service for artists without powerful local hardware, using cloud-based GPU servers.
How it works:
1. Install Glaze, including necessary machine learning libraries and pretrained models.
2. Input single or multiple images into the Glaze application.
3. Adjust intensity and render quality settings.
4. Use the preview function to assess the estimated final result.
5. Run Glaze to apply the style cloak and save modified images.
Integrations:
Amazon AWS, PyTorch
Use of AI:
Glaze employs adversarial machine learning techniques to create perturbations that disrupt AI training. These perturbations are designed to minimally impact human perception while significantly affecting AI models' ability to learn and replicate the artist's style.
AI foundation model:
The tool is particularly effective against popular generative models such as DALL-E, Midjourney, and Stable Diffusion.
Target users:
- Artists at risk of having their work mimicked by AI models
- Creative professionals including musicians, authors, journalists, voice actors, dancers, and choreographers
How to access:
Glaze is available as a desktop application for Windows and MacOS, as well as a web service (WebGlaze). It is not open-source to maintain the integrity of the tool and raise the bar for adaptive attacks.
Applicable Industries:
- Art and Design
- Media and Entertainment
- Publishing
- Advertising
-
Supported ecosystemsAmazon, Apple, Google
-
What does it do?AI Art Protection, Style Preservation, Digital Rights Management, Creative Content Protection, Adversarial Machine Learning
-
Who is it good for?Graphic Designers, Creative Content Creators, Visual Artists, Digital Illustrators, Fine Art Photographers