- Aug 2, 2017
- 714
- 2,802
Oh that's pretty neat to add new "object" on Facedetailer. I'll have to study more workflow for Comfy.Facedetailer also can be used to add an object if needed, but the area is limited to face, and there's only one kind of object I added to face so far, you know, lol.
Yes, I got several frustations when working with loras, especially if I need new lora to achieve the context of the picture, since it usually have effect on the style. Tweaking the weight takes time, especially with my slow-ish gpu, lol.
Yeah adding new loras for pose can sometimes add styles, which could be annoying. Controlnets would be most accurate, but can be time-consuming.
I wonder if you could train your own Lora Style if you generate enough variety of images with the mix lora weight you have. You need a minimum of 40 images up to 100 images (40 is fine, but don't go over 100) of various areas with or without people. You could train it in the
You must be registered to see the links
for 500 Buzz, which is nothing compared to the new 2000 Buzz for Flux (I wonder how SD3.5 is now).I've only trained a Lora locally with Flux Gym, which was pretty neat, but the Civitai Trainer is way better quality. Your Lora would technically be called a Lora Merged. Let me know if you need help