Used AbyssOrangeMix2 with Anything V3
Much better results than without it, I used it between 0.3 and 0.7.
Works great with other LoRAs
Description
Retrained with higher dimensions, better quality imo
FAQ
Comments (8)
In your retrained version, did you change the trigger words (concepts)? I really liked that "restrained" and "bound arms" were in there in the first version, since that's what I really want from a model.
No, they are the same!
@halfawave Interesting. I'm dabbling with your first and second version now, with a realistic model called Deliberate. The second version pulls the images much, much further toward an anime style, whereas the first version mainly affects the concepts themselves without having too much of an impact on the style, colour, background, etc. I guess that's a consequence of V2 having more parameters, so it's a better learner so to speak. It's picked up on a lot of things that weren't part of its explicit learning goals.
@kwik that is good to know, I'm making a lora for doggystyle now, I will make some versions with different network dimensions to see what effect it has
The hole above the ball gag I think is very difficult to train
What's the meaning of gag?
Gag is the thing in their mouth
A Joke :D
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.