Pushing Limits

Experiments in Stable Diffusion

How realistic can the output sdxl generates be at 4-8k? Pretty darn close 😀
Generation with DPMPP 3M SDE sampler, refined with Euler sampler (both using Karras scheduling).