Stylebreeder: Exploring and Democratizing Artistic Styles through Text-to-Image Models
Matthew Zheng, Enis Simsar, Hidir Yesiltepe, Federico Tombari, Joel Simon, Pinar Yanardag
2024-06-24
Summary
This paper introduces STYLEBREEDER, a large dataset and platform that helps people explore and create diverse artistic styles using text-to-image models.
What's the problem?
As technology advances, creating digital art has become easier, but many artists still face barriers to accessing tools that allow them to express their creativity. Traditional methods of art creation can be limiting, and not everyone has the resources or skills to produce high-quality artwork. There is also a need for a better understanding of different artistic styles and how they can be combined or personalized.
What's the solution?
The authors created STYLEBREEDER, which includes a dataset of 6.8 million images and 1.8 million prompts generated by 95,000 users on the Artbreeder platform. This dataset allows users to explore various artistic styles and generate personalized content based on their interests. The research highlights unique styles that go beyond common categories, providing insights into the collective creativity of users worldwide. They also developed a style atlas to help users navigate different styles and made their models available for public use, encouraging more people to engage in digital art creation.
Why it matters?
This research is important because it democratizes access to artistic tools, allowing more people to create and experiment with digital art. By providing a platform for exploring diverse styles and personalizing content, STYLEBREEDER fosters a more inclusive artistic community. This not only enhances individual creativity but also contributes to a broader understanding of art in the digital age.
Abstract
Text-to-image models are becoming increasingly popular, revolutionizing the landscape of digital art creation by enabling highly detailed and creative visual content generation. These models have been widely employed across various domains, particularly in art generation, where they facilitate a broad spectrum of creative expression and democratize access to artistic creation. In this paper, we introduce STYLEBREEDER, a comprehensive dataset of 6.8M images and 1.8M prompts generated by 95K users on Artbreeder, a platform that has emerged as a significant hub for creative exploration with over 13M users. We introduce a series of tasks with this dataset aimed at identifying diverse artistic styles, generating personalized content, and recommending styles based on user interests. By documenting unique, user-generated styles that transcend conventional categories like 'cyberpunk' or 'Picasso,' we explore the potential for unique, crowd-sourced styles that could provide deep insights into the collective creative psyche of users worldwide. We also evaluate different personalization methods to enhance artistic expression and introduce a style atlas, making these models available in LoRA format for public use. Our research demonstrates the potential of text-to-image diffusion models to uncover and promote unique artistic expressions, further democratizing AI in art and fostering a more diverse and inclusive artistic community. The dataset, code and models are available at https://stylebreeder.github.io under a Public Domain (CC0) license.