#weight: -3.0 to 3.0 #positive: more weight #negative: less weight V2: It is more capable on both ends of the spectrum but the range you give the lora is less. I uploaded a few days ago but something messed up and I couldn't update it. I have been working non stop on my training algorithm and it is becoming far more stable while affecting the case concept of the character and the scene less and less. With this one, you get almost no artifacts with contrast and scenery as you scale and you can get much more extreme results, leaving more in the middle.