Weight 0.7-1
V2
Trained on Ahri's new model and should be easily customizable. I added many different style loras below to showcase this.
There is a slight bias for some poses to make her ears white, if that happens you can put "white ears" in the negative prompt.
Adetailer/face segmentation is not needed but overall helps with the face quality.
V1
Trained on Ahri's original design, before the small visual rework. Add/remove braid for preference.
If using multiple other Loras use lower weight.
Square: 1024x1024
Portrait: 832x1216
Landscape: 1216x832
Description
FAQ
Comments (3)
I'm having trouble recreating the first image. I used your lora and copied in all the settings using autismmix (confetti) exactly the same as the image but it just looks different.
DPM++ 2m Karras,
25 steps,
2048x2048,
7 scale,
380488927 seed,
clip skip 2,
Exact same positive and negative prompt, with same lora weight 0f 0.8.
Was this image actually generated using these settings and this version of the lora/autismmix model?
i've experienced that using a different app (be it a1111, comfy, swarm) or even just diferent versions of the same app can lead to pics not exactly being the same
Hi I know this was commented long ago, but it's probably what Fuuka said as well as the lora used for this image was when I was figuring out which lora trained was working best at the time as it has "lora:AhriIL-07:08" which should probably be "lora:AhriIL:08" for u as when it's downloaded I believe it losses the number following