Type | |
Stats | 654 1,983 19 |
Reviews | (107) |
Published | Jul 11, 2023 |
Base Model | |
Training | Steps: 5,800 Epochs: 2 |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 B58C79D6B9 |
This model creates a "fae" or fairy inspired style, which as it gets more weighted becomes more intricate and colorful. It tends towards luminous, vivid, and manyfold colors - a friend called it "Mardi Gras" style, which isn't all wrong - and accessories such as wings, flowers, and animal ears. Faestyle doesn't produce great hands or feet by itself, but the eyes are sometimes quite decent. It meshes well with urban and cyborg concepts, and likes to decorate skin, including on the face. No faeries were harmed in the making of this lora.
The examples I am including to start with do not use any additional loras or embeddings beyond this lora and the checkpoint model. On the checkpoints I tested, the sweet spot weight for V2 is about 0.3 to 1.4 and for V1 is about 0.6 to 0.9 inclusive. Importing at 0.7 in your prompt would look like this: <lora:faestyle:0.7>. No trigger word is required.
It's my second uploaded model, and I'm not convinced it is great, but it often creates images I enjoy. V2 is so much better after just a few days, so I'm learning! Comments and feedback greatly appreciated!