lllyasviel/LayerDiffuse

Why not just ControlNet style for conditional generation?

xiankgx opened this issue · 0 comments

Seems like the alternative architecture for conditional image generation is SD inpainting channel concat style. Why not use Controlnet style instead?