Stable Diffusion XL

Stable Diffusion XL, or SDXL, was the successor to the original Stable Diffusion line and was designed to deliver higher resolution, better coherence, and more realistic detail than earlier releases like 1.5. It introduced a dual stage architecture with a base model and a refiner, improved text handling, and more accurate anatomy, which made it a notable technical step forward for open diffusion systems. SDXL generates images at a native 1024 by 1024 and played a central role in the maturation of community fine tunes, LoRAs, and downstream tooling. Although it is no longer the state of the art compared with modern transformer first models, SDXL remains an important milestone in the evolution of open source image generation and is still used as a baseline.

July 26, 2023

Stability AI

Text to Image (T2I)