1 paper
Lidia Troeshestova, Alexander Ustyuzhanin, Sergey Kastryulin
Recent diffusion editors perform diverse instruction-based edits while conditioning on the source image at every denoising step. Yet persistent source-image conditioning can limit…