|
--- |
|
license: openrail++ |
|
language: |
|
- en |
|
tags: |
|
- anime |
|
- image generation |
|
- PonyXL |
|
- Stable Diffusion |
|
- SDXL |
|
--- |
|
|
|
## MikuFactorXL |
|
|
|
![image/png](https://cdn-uploads.huggingface.co/production/uploads/6342619a9948f573f37a4a60/Y19X7ZFSBFd9unBCWY2h3.png) |
|
|
|
## About |
|
|
|
This is my first PonyXL merge, it is just based off random models I picked off CivitAI. The results are amazing to me at least, best PonyXL model avaible I think personally! However I am very biased. I tried to select models for following |
|
|
|
## Model selection criteria |
|
- Good small details |
|
- Ones that look least like AI slop, more solv |
|
- Ones with least artifacts in example images |
|
- Ones that I find personally appealing |
|
- Bright colors |
|
|
|
## Model merge techique |
|
|
|
Idk I just merged bunch of models in ComfyUI, they are all listed in `mega-merge.json` to look at good luck finding them all on CivitAI. My PC almost crapped itself and had to make 200GB swap file, so like would be nice |
|
|
|
## Observations |
|
|
|
- Great quality |
|
- Amazing prompt following |
|
- Good annotomy most of time, can still make mistakes |
|
- Very creative posing |
|
- Overall very creative |
|
- Faces have sovl |
|
- Good small details |
|
- Thin line art |
|
- Understands a lot of concepts |