Reference from Azur Lane Character, Enterprise .

Pony is running under different base then usual SD format quality prompt, please look into Ponydiffusion page checkpoint or refer to below instruction.

If you feel your VRAM are not up to SDXL gen, consider try out SD Forge:
https://github.com/lllyasviel/stable-diffusion-webui-forge
Is a improved and way better A1111 which giving you an edge to use SDXL gen with lower VRAM

Simplify about Model:

• default weight is recommended

• Hires and Adetailer a recommended

• Quality Prompt ``score_9, score_8_up, score_7_up``

• Avoid using negative prompt as much possible, it could affecting the quality of your image.

• Bon voyage contain extremely less dataset, may seeing as overfit

• The lora are train with cropped image, sd possibly generate cropped gen.

• consist of 9 costume.

• Recommend to run with IncursioMeme(Pony)/AutismMix(Pony) and Ponydiffusion

Below are test and preview between non-style and self make Lora

Default:

Heroic Finery:

Wind Catcher:

Reindeer Master:

Blooming Peony:

Diving Under Blue Skies:

Anniversary Ride:

Starlight Oath:

Bon Voyage:

About Image Gen:

• The image preview are generate with forge webui.

• All Image are produce with Model IncursioMeme(Pony).

• All Image are gen with Hideo Minaba Fraction Style Lora.