Stable Diffusion 1.5

Stable Diffusion 1.5 is the legacy open source checkpoint that defined early community image generation. It has been surpassed by newer architectures with better coherence, detail, and fidelity, but it remains historically important because it powered the first wave of community fine tunes and helped establish the ecosystem of tools that creators still use today. Native output runs at 512 by 512, with higher resolutions achievable through tiling and upscaling. Although it is no longer considered state of the art, SD 1.5 continues to be referenced as a baseline in research and benchmarking since it represents the classic latent diffusion era before the major advances in training quality, preference tuning, and visual consistency that arrived with SDXL, FLUX, and the modern transformer first generation of image models.

October 20, 2022

RunwayML

Text to Image (T2I)