v3 gives much better results and works well at 1.0 strength. It is trained on AnyLora.
v2 is trained on NAI and gives much less coherent results. If you want to use it anyway, then I recommend putting strength up to 1.4.
Description
-Improved dataset
-Now trained on AnyLora
FAQ
Comments (3)
Since I do not quite understand if people can even see the text I attached to the image post below I will write it here once more in hopes of finding a solution:
"After v2 already worked wonders I was very excited to try out v3 but for some reason all my generations come out as overcooked while also destroying the style of the Model and Lora used. My attempts at fixing it by either lowering or increasing the strength of the paf Lora did nothing to improve it. I even retried the same prompt with v2 and it worked fine.
Do you have any idea what could cause this? The example uses v3."
Guess it clashes with the style lora and v2 is a bit more flexible with that. I got some working results with it though with lower strength in the style.
I can't get this to work no matter what I try.