Comment on VMix: Improving Text-to-Image Diffusion Model with Cross-Attention Mixing Control
p03locke@lemmy.dbzer0.com 4 weeks ago
It almost seems like it’s forcing a certain kind of lighting without prompting. I wonder how well it performs when specifically given something that isn’t “natural light”.
Even_Adder@lemmy.dbzer0.com 4 weeks ago
It’s an adaptor like LoRA. You train a different one if you want another aesthetic.