ZIT 5.1 Version is to fix ethnicity change and over saturation.
Illustrous and ZIT Version's are safetensors.
Zıt version works between -4/+4 without breaking.
It is a slider lora negative values are for "skinny" and plus values are for "voluptous". Works best between -2/+2 but don't hesitate to experiment. Negative values desaturate and positive values higher than 2 saturates the model. I am using Sliders to train these If you have any suggestion to prevent saturation change I am more than happy to hear.
Description
FAQ
Comments (8)
Safetensor version?
If you know a method, I can happly convert it.
Something's coming up in the part-- the clip. There's examples at the bottom. If I set the model to a negative value, and the parameter - clip to a positive value, the color appears. Here congechno there are assumptions and the picture itself is not for beauty, additionally affecting the color is HDR and sharpness. Also have to overestimate the config. And in this case there is a movement to equalize the color.
For understanding this is the settings in Comfi in Lora.
Safetensor version?
I wasnt able to find a way to convert it to safetensors.
What? Only 9MB???? Moreover delivered as an insicure .pt file?
It is 9 mb because it is a slider Lora, it is pt because the script I used was only giving pt output. I switched to flux so I dont care about converting it.