Please support the artist and do not use this model for monetary gain!
I'd appreciate it if you could upload pictures here if you use this model, I'm curious to see what others will do with it, and if I like it, I will tip buzz! Thank you!
Anima LoRA
My dumb ass didn't know that the base checkpoint already learned this style before doing training. But now I guess it will know e621 furry knowledge and does better furry characters in the artist style (especially the head). From my testing, a bit better at NSFW too compared to just using the artist tag alone without LoRA.
Also, I uploaded both epoch 4 and epoch 2 (in Optional), because I couldn't decide which one's better. Epoch 4 should be better generally but use epoch 2 if you want a lighter touch.
I hope you like it! Trained using Danbooru tags. Does good both furry, human.
Usage Guide
Trigger Word: @traver009
Recommended Weight: 1.0
Preview Image Parameters (preview images use epoch 4)
Checkpoint: Indigo Furry Mix Anima v1.0
Sampler: ER SDE
Scheduler: SGM Uniform
Steps: 30
CFG Scale: 4.0
Hires.fix:
Upscaler:
4x_foolhardy_Remacri(1.5x scale)Denoising Strength: 0.3
Steps: 15
*These settings are by no means the best, just what I personally use, but if something looks off on your generation, you can maybe give these a try!
**I did some inpaints on preview images, so if you are trying to replicate through txt2img only, it might not look 100% the same!
Description
Training data from e621 and danbooru.