Type | |
Stats | 722 523 2.3k |
Reviews | (47) |
Published | Oct 16, 2024 |
Base Model | |
Training | Steps: 818 Epochs: 10 |
Usage Tips | Clip Skip: 1 |
Hash | AutoV2 11453A97FF |
This LoRA adds seated soles to FLUX.
Men and women, seated or lying down, shot from a low angle depicting the soles of their bare feet, socks, black stockings, or sneakers (generic, nike, and converse all star). One sole or two.
LoRA nerds: I didn't take out the watermarks on this, it didn't have many, but it only gives them like once out of 20 times. Maybe if they are in every image or like half the images (This one is 44%) in your dataset it becomes a problem? I'm leaving them on in all my LoRA from now on until I start running into them more. I did not add them to the captions and took it out when they got caught that one time by the auto captioner.
1024x resolution is the key to good FLUX LoRA made onsite, that or Prodigy optimizer.
FLUX lora just do not work on Civit's generator, they barely work on my local, Tensor art beats both my 16gigs of vram and whatever Civit has. I get maybe one good gen out of five over there with this, one gen out of ten with NSFW. It's over here.