
So, this is a weird one. This LoRA has two versions, one trained off the original showcase, and another that had some additional images thrown in of the same style to help with the training.
I would suggest using With Additional Images since it produces better images while still being faithful, but I put both up for the sake of personal preference.
On Counterfeit 3.0, these LoRAs are very strong, to the point that they still work at 0.2 weight, going above that can produce errors, but might be necessary.
On other models, I would suggest using this at weight 1, the style won't be the same but it will have a strong effect.
This loRA is dedicated to ppokket for complaining that Counterfeit was a scam since nobody could make images of a style similar to the original showcase. In all seriousness dude, you inspired me to make this to solve that problem.
描述:
训练词语: Counterfeit Style
名称: CounterfeitStyle2.safetensors
大小 (KB): 9330
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success