ModelsLab/AingDiffusion - V10

Major mix rework, better overall foreground performance.

Background sometimes work flawless, sometimes is a mess. (just add more negative TI to fix this lol)

Changing default VAE to kl-f8-anime2 since it works better on most use case.

Comparison between v10, v9.2, and v8.17 is here

===

Personal note:

Putting Stable Diffusion on SSD is really helping the whole development process.

I believe a new GPU also gonna help. My RTX 2060 is kinda slow and cannot keep up with SDXL, that's why I'm stuck with SD 1.5. You can help me buy a new GPU here at ko-fi