Type | |
Stats | 324 1,230 |
Reviews | (48) |
Published | Jun 17, 2024 |
Base Model | |
Trigger Words | maronk |
Hash | AutoV2 98BA6A7710 |
This is my first LORA, so it's not perfect. I will improve the dataset later on. ❤️
Recommended settings
Pony
Weight: ~0.8 is ideal.
Works only for Pony-based models, I recommend using AutismMix.
To achieve the same results, don't forget to add the same positive and negative quality tags as I do in the sample images.
SD 1.5
Weight: ~0.7 to 0.6 is ideal.
Add 1990s \(style\) to enforce the typical 90s anime look.
Use masterpiece, best quality in the positives at the beginning of the prompt, and (low quality, worst quality:1.4) at the beginning of the negative prompt.
Trigger prompt
Pony
maronk,
SD 1.5
maronk, doughnut hair buns,
Additional info (SD 1.5)
This LORA was trained on AnyLoRA, so it is compatible with most SD 1.5 anime models.
I noticed there aren't many high-resolution pictures of Maron available online, which might have limited the quality of this initial LORA attempt.
If her face needs refinement, consider using inpainting with a strength of 0.2-0.25 on the masked area (the regions requiring touch-up). I've had good results using this approach.
Also, sometimes, her double buns are inconsistent, and the school uniform needs better training with more pictures, although you can probably achieve a similar look by following the keywords.