This is a playground, where each version is a different lora.
These are unfinished experiments that still prove very useful from time to time, when I combine loras to make something glorious.
- All the models are trained by me. If the model starts with "pin" - means it's dataset only contained trained on my own genrations and edited work; if it starts with "pon" - means it either was trained only on artist's work or also contain artist's work.
-- I mostly use MugenMalu mix 4.2 checkpoint as base, so I'd suggest you try it out too.
--- I also use ComfyUI - no problems there thus far.
A1111 though - if you add it as lora into positive - seems to almost work ok(except generation takes longer); if you add it via "additional networks" - you may encounter "silent" error, the effect of which is lora doesn't seem applied at all, no matter the weight. :(
Also, on tensor.art it's not working yet - trying to figure out why along with their support.