These outputs were created using our base SD3.5 Medium model, without upscaling or refinement. Clownshark also used the “ClownSampler” from the RES4LYF repo, which includes nearly 30 unique sampler and noise modes. https://
github.com/ClownsharkBatw
ing/RES4LYF
… (2/3)

→ View original post on X — @stabilityai