EyeControl - Jitome: v2: EyeControl_Jitome_v2.safetensors

Expanded training set. Less impact on style. 5000 steps @ batch size 4 using DAdaptation preprint.

⭐ If you make something neat, please share. I don't want your money but I like feeling useful. ⭐

For some reason, SD doesn't know jitome eyes (flat top eyes).

v2 is up. Larger data set, and less impact on style.

Weight range from 0.8-1.5 should be good depending on model. expressionless or unamused for staring. empty eyes for flatter shading on the eyes. grin or anything happy needs higher weight on the lora.