Version 2 of the cheatsheet released. It can be found here: https://civitai.com/articles/5473
I have been messing with Pony for a while an notice many people creating lots of foggy, messy, images. So I have created a nice little base line for several styles/sources.
Please read notes for each type for any extra info other than the defaults.
If you have any suggestions, improvements, or request feel free to ask. However keep in mind I do not do artist styles.
Each image (unless otherwise stated in notes) was made with the follow info:
Model: https://civitai.com/models/288584?modelVersionId=324524
Prompt: I used the base and "1girl, face" as the prompt to keep things close to the same.
Sampling Method: DPM++ 2M SDE Karras
Sampling Steps: 35
Width: 1024
Height: 1024
Seed: 2242263340
Update 4/19/24:
A good alternative model is: https://civitai.com/models/356456/matrix-hentai-pony
I found this gives a bit more power to the styles and characters. I am still working on testing with settings and such and will update more after testing. I have also found a few new styles I will be adding as well
2.5D:
Notes:
Lora: https://civitai.com/models/297619
Stength between 0.5-0.9 to get more or less of the 2.5D look
Prompt:
(<lora:Realistic_2.5DAnime_Merge1:0.8>, score_9, score_8_up, score_7_up), BREAK
1girl, face
Neg:
(score_3_up, score_4_up, score_5_up)
Preview:
80's Anime:
Notes:
Lora: https://civitai.com/models/340679/80s-anime-woman
The image is pretty boring but I didn't want to change the prompt to keep things the same. But adding simple background to neg normally fixes the backgrounds.
Prompt:
(score_9, score_8_up, score_7_up, 80swa, <lora:80s_Anime_Woman:1>), BREAK
1girl, face
Neg:
(score_3_up, score_4_up, score_5_up)
Preview:
90's Anime:
Notes:
Lora: https://civitai.com/models/333041
Prompt:
(score_9, score_8_up, score_7_up, Vintage, 1990s \\(style\\)), <lora:OldAnimeStyle_XL_v3:0.75>, BREAK
1girl, face
Neg:
(score_3_up, score_4_up, score_5_up)
Preview:
Pixel Art:
Notes:
Lora: https://civitai.com/models/43820
Prompt:
(<lora:pixel:1> , score_9, score_8_up, score_8, pixelart, pixel), BREAK
1girl, face
Neg:
(score_6, score_5, score_4)
Preview:
Realistic:
Notes:
Model: https://civitai.com/models/365041?modelVersionId=407890
Does really good with bodies, faces could still use work. I normally create something then use adetailer to fix the faces with a normal SDXL realistic model
Prompt:
(score_9,score_8_up,score_7_up,score_6_up), BREAK
1girl, face
Neg:
(score_5,score_4)
Preview:
3D:
Notes:
Lora: https://civitai.com/models/331812
Prompt:
(score_9, score_8_up, score_7_up, 3d, <lora:RetouchXL_PonyV6_v2:1>), BREAK
1girl, face
Neg:
(score_3_up, score_4_up, score_5_up)
Preview:
Source:
Notes:
Lora: https://civitai.com/models/335860/x3d-style-pony
What is the difference between this and 3D?...More Shiny?
Faces don't do great, adetailer with the same model normally fixes this
Prompt:
(score_9, score_8_up, score_7_up, X3dCE, 3d, <lora:X3dCE_V2:0.6>), BREAK
1girl, face
Neg:
Preview:
Anime:
Notes:
Feel like this could be better. Going to work on improving this.
Prompt:
(source_anime, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up), BREAK
1girl, face
Neg:
(score_3_up, source_furry, source_pony, source_cartoon)
Preview:
Furry:
Notes:
Yes, just cat ears. But wanted to keep with everything else.
Prompt:
(source_furry, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up), BREAK
1girl, face
Neg:
(score_3_up)
Preview:
Cartoon:
Notes:
Prompt:
(source_cartoon, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up), BREAK
1girl, face
Neg:
(score_3_up, source_anime, source_pony, source_furry)
Preview:
Pony:
Notes:
Prompt:
(source_pony, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up), BREAK
1girl, face
Neg:
(score_3_up, source_anime, source_furry, source_cartoon)