I prohibit companies like PixAI, Tensor, etc. and their users from uploading my checkpoints to websites outside of Civitai. I will do it myself if I want!
ā Buy me coffee, if you want to support me: https://ko-fi.com/tzigoai
This is my first model and it's far from perfect, but I'm working on it.
Please rate, comment and post your results with my model! Thanks!
__________________________________________________________________________________________________
V1.042_VAE_Hyper
Sampler: any, but I recommend Euler A because of the speed. And that's what Hyper is all about.
CFG: 5 - 8
Steps: ~12, 6 steps worked in the test, with the corresponding loss of quality
__________________________________________________________________________________________________
V1.042_VAE
Now based on anima_pencil-XL and therefore Animagine XL V3.1.
This should also make most of the characters included in Animagine work, but I haven't tested it further.
It is now more detailed and has improved eyes and better hands (mostly). It is also less 2.5D now. At least I was satisfied with my tests.
Sampler: Euler, Euler a or DPM++ 2M Karras worked well in my test, but any other should work too.
CFG: 7~10
Steps: ~30, also works at ~20
Merge recipe:
1 animaPencilXL + (FiaMix XL - Animagine XL V3.1)*0.75, Add difference(trainDifference)
2 [1] + (Rae Diffusion XL - Animagine XL V3.1)*0,65, Add difference(trainDifference)
3 [2] + Modern Disney XL*0,35 Weight sum(normal)
4 EnvyMegaMixXL01 + (EnvyAnimeXL - DreamShaper XL)*0,8, Add difference(trainDifference)
5 [4] include Loras
6 [3] + [5]*0.42 Weight sum(cosineA)
7 Pilgrim_Unreal_SDXL + (Pilgrim_Midjourney_SDXL - sd_xl_base_1.0)*1, Add difference(trainDifference)
8 xVision XL + [7]*0,48 Weight sum(cosineA)
9 [6] + [8]*0,37 Weight sum(normal)
10 [9] +(Quiet Goodnight XL -AAM XL(Anime Mix))*0,63, Add difference(trainDifference)
11 Shambhala.XL include Loras
12 [10] + [11]*0,24 Weight sum(cosineA)
13 RealCartoon-XL + Yamer's Anime*0,52 Weight sum(cosineB)
14 [12] + [13]*0,365 Weight sum(normal)
15 [14] + CheemsPuff XL*0.38 Weight sum(cosineA)
16 [15] include Loras + Baked VAE_ sdxl_vae_fp16_fix
Credits to the respective authors.
__________________________________________________________________________________________________
V0.63B Update:
I completely recomposed the merge and still tried to keep the typical style.
Now included are:
As well as a few of my favorite Lora's.
Credits to the respective authors.
__________________________________________________________________________________________________
It's a 80% merge of TheCheckeredSausage's awesome CheemsPuff XL and 20% KandooAI's Juggernaut XL for some realism.
Credits to the respective authors.
Sampler: Euler, Euler a or DPM++ 2M Karras
Steps: 30 and more if necessary
CFG: 7-10
For LCM+Turbo version:
Sampler: Euler a required
Steps: 6-8
CFG: 1,5 - 2,5