Type | |
Stats | 1,228 11,273 |
Reviews | (184) |
Published | Mar 30, 2024 |
Base Model | |
Training | Epochs: 5 |
Usage Tips | Clip Skip: 2 |
Trigger Words | smr03, pointy nose |
Hash | AutoV2 C46B728486 |
This is art style is inspired by the works of an artist known as Samara03. Unique and original style with angular drawing and strokes. LORA was trained on Pony Diffusion V6 XL/Anything-v3.1 model (with vae from SD). Suitable for both t2i, i2i and mixing styles as well.
For PonyXL
there are two versions OG and v1.0
OG - keeps the art style as close as possible to the original.
v1.0 - creates a style inspired by the author's recognizable handwriting, but not exactly the same.
The recommended weight is 0.7-1.0
0.5-0.7 is for light stylization, 0.9-1.1 for hard stylization.
Recommended trigger words: smr03, ponity nose
Also for v1.0: you may need to add "horns" tag to the negative, or remove "smr03" tag from prompt, if you keep getting horns in your results.
My optimal settings:
CFG: 5-7
Steps: 30
Clip skip: 2
Sampler: Euler a
LORA weight: 0.7-1.0
Use Adetailer and Hires.fix for better results!
And have fun! Kva!
For 1.5
The recommended weight is 0.7-1.0
Recommended word: angular drawing, smr03, pointy nose, sketch
Also: you may need to add "glasses" tag to the negative ¬_¬
My optimal settings:
CFG: 5-7
Steps: 20-25
Clip skip: 2
Sampler: DPM++ 2M SDE Karras
LORA weight: 0.7-1.0
Use Hires.fix for better results!
And have fun! Kva!