Type | |
Stats | 795 42 |
Reviews | (106) |
Published | Mar 20, 2023 |
Base Model | |
Trigger Words | ilulu_\(maidragon\) |
Hash | AutoV2 0BBF2A8784 |
Trigger word: ilulu \(maidragon\)
Recommended weight: 0.8-1.0
Useful descriptions: pointy ears, low twintails &| twintails, colored tips, red eyes | orange eyes, slit pupils, sharp teeth, yellow horns, dragon tail & dragon girl, shortstack | petite
Character can be generated without additional descriptions just fine, but in case you don't get what you want try using them in your prompt.
v2-Epoch11: should better draw her twintails and horns in cases where it doesn't. Updated Data Set: added some missing poses tags, "dragon girl" for pictures where there is "dragon tail", and removed Hair Length and Hair Color (with exception of "colored tips").
Epoch8: may better depict the character, but may start losing prompt details.
Epoch5: may follow original image better, but may not get some character features as good.
This lora was trained on 80 different pictures of the character with 3 repeats. I was trying to train it on many different models and in the end, between Anything-v3 and NAI, decided to land on Anything-v3 as it seemed to produce better results. Data with her dragon hands seems to have made Lora worse, so I decided to exclude them. All images shown were used with different models at either strength 0.8 or 1.0.
The models used for preview images are: MeinaMix v8, Grapefruit v4.0, Color Box Model, DualPersonality - Normal Line, Animelike 2.5D, GuaMix3;