Type | |
Stats | 339 619 265 |
Reviews | (78) |
Published | Apr 7, 2024 |
Base Model | |
Training | Epochs: 10 |
Usage Tips | Clip Skip: 2 |
Trigger Words | bodychan bodykun grey skin wireframe |
Hash | AutoV2 C99DAB582B |
Finally, LoRA for helping artists!! Maybe not...
Recommended prompt: bodychan or bodykun ,
grey skin, (wireframe), (faceless, no eyes), bald
Both versions which are AnimagineXL3 (SDXL1.0) and Pony Diffusion XL 6 (recommended AutismMix SDXL), but both has their advantages..
I trained the LoRA separately for the LoRA mixing compatibility. AnimagineXL3.1 if I'm pretty sure can do ControlNet stuff (haven't tried that yet), while Pony does neat at 3D (no idea about the other things idk lol), however i'll be talking about the problem later..
Both can do various poses, for example AnimagineXL3 version can do T-poses, while Pony can do jack o pose.
It is RECOMMENDED to use ControlNet if you're gonna do pose references, Regional Prompter or Latent Couple too for multi characters and maybe others.
ANOMALY ALERT! AN ANOMALY HAS DETECTED!!!!!
When using the Pony version of the LoRA and trying to generate multiple characters with different gender, there is the most chance an ugly bastard will appear. Idk what the problem is but I need alot of cherry picking to not get the ugly bastard and actually get both bodychan and bodykun.. It also ignores the grey skin and uses the humanoid skin tone instead. Woah atleast AnimagineXL3 version of the LoRA doesn't do that!!
Have fun with it
Datasets: https://drive.google.com/drive/folders/1FBUP2dZw-3GG-BXn_1_woSGLjSaIVctO?usp=sharing