Bria Launches Open-Source Text-to-Image AI Model

Bria has released its third-generation open-source text-to-image AI model, which matches industry-leading performance with only 4 billion parameters, announced in a press release. The model, available on Hugging Face, is built entirely on licensed data, ensuring compliance with copyright laws.

The new model is designed to be twice as easy to fine-tune compared to other leading models, requiring 50% less compute and data to achieve similar results. This efficiency is achieved while maintaining high performance, as demonstrated in blind preference evaluations conducted among thousands of design professionals.

Bria's approach focuses on ethical AI development by partnering with artists and content creators, ensuring fair compensation and eliminating copyright risks. The model is part of Bria's platform-as-a-service offering, which includes features like content moderation and enterprise security standards, supporting multiple access methods such as plugins for Figma and Adobe Creative Suite.