
v2. using new tagging, and training, No need for trigger words + she is now much freer than just stuck with what was in the video.
Without trigger words, it is much easier to change and get rid of the cloth.
But if you can't get her close to CG, you can use "owsombra" to trigger it. You may need to lower the weight while using the word. And it is very very hard to go nsfw with the trigger word.
All v2 image is made on my own checkpoint "GoodFit" This checkpoint is made for Lora like this, and it will give you better results. Other checkpoints might work, but not tested.
Hope you like it :)
share some post & join the discord if you like :D
-
-
-
-
-
v.somethingdifferent-- I won't call this close to CG-like, and I have no idea how I trained this version. changing background, changing clothes as much as you can think of, other than she looks thicker than the she in the CG, don't know if it is good or bad, If you like a cuter version or a more CG-like version use v1.
I did all my images around 0.65 weight, you can start from there.
v.1---------------
I think am getting better at making Lora. This is my 4Th one
Love those overwatch CGs, so this is also one of the CGs like Lora made from "Infiltration"
This Lora for me is well-trained, making a pose, changing faces, clothes, background, etc.
But still has a few bad things, for some reason, it keeps getting from the above views, not sure what cost that, and also not very CG-like, but it is one of those well-fit Lora. I'm definitely going to make a V2 CG-like version.
For V1 version weight, 0.5 to 0.75 should give you great results.
描述:
训练词语:
名称: owsombrav2p.safetensors
大小 (KB): 295027
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success