Verified: 2 years ago
SafeTensor
Type | |
Stats | 292 |
Reviews | (52) |
Published | Aug 3, 2023 |
Base Model | |
Training | Steps: 2,000 Epochs: 2 |
Trigger Words | ruinedshyvana |
Hash | AutoV2 E4009E89AF |
A LoRA mainly trained on Ruined Shyvana's in-game assets (model, splash art, Ruination event sprite). It was trained on NAI model, but well, it works better with semi-realistic models like JAM.
The model often fails at hands, has about 20-30% chance to mess up the face when used on STR 1, pretty much always fails the eyes (you can probably see it on the examples).
Dragon form or wings weren't trained at all, her weapon was in the dataset, but I have no idea if it works.