Version 2 optimizes brightness transition issues and enhances stability. The weight range has been expanded to -10 to 10, with more pronounced effects starting at 6 and -6.
v2版本优化了亮度变化的问题,提高了稳定性
权重范围扩充为-10~10,6和-6开始会有比较强烈的表现
This LORA can slide between anime (negative weight) and real (positive weight). The LORA range is -5 to 5. Excessive weight will cause brightness changes: anime becomes brighter, real becomes darker. For image editing, use "replace style" or other similar prompt.
这个lora可以在anime(负权重)和real(正权重)之间滑动
lora范围在-5~5之间,过高的权重会导致亮度变化,anime会变亮,real变暗
如果是图片编辑下,可以用“replace style”或其它类似的意思作为提示词
Description
FAQ
Comments (4)
better results most of the time with v1. both great and work really well, but with a face reference image v1 is so damn good.
Got it, but I must remind you that v2 requires a higher weight to take effect, as I have adjusted the training scale. I am also planning to make a v3 version, and try to control the weight to less than 2 for it to take effect.
@yogurt7771 ya i pump it up to 8/9 for v2 and its okay, it still loses facial features at that strength. V1 i can keep at 2 and it works pretty flawlessly. V2 fixes body part errors better though and the lighting is usually better, whereas v1 falls flat in most images. Its a hard choice to make! lol
This is way better than other A2R loras on here. As usual you gotta dig past the popular for the best stuff!

