Type | |
Stats | 444 1,023 |
Reviews | (71) |
Published | Nov 10, 2023 |
Base Model | |
Training | Steps: 2,000 Epochs: 8 |
Usage Tips | Clip Skip: 1 Strength: 1 |
Trigger Words | Inteleon |
Hash | AutoV2 0B2F5B02C9 |
Inteleon LoRA trained on NovelAI with Kohya training scripts using the DAdaptAdam optimizer at 512x512 resolution. Inteleon's skin has a lot of different colors all very close together, expect them to get mixed up a bit. Adding (blue torso, black thighs, black hands) to my prompt improved consistency a lot. Prompting for a skinnier build seems to improve consistency, which makes sense given how twiggy Inteleon is. Doesn't always include Inteleon's "cape". The black thighs can turn into clothing if you're prompting for things like lingerie or a bikini, I recommend seeing what kind of clothing the thighs turn into and adding it to your negative to fix that. I had to add (thigh highs) to my negative to get the lingerie pic in the examples. Should have wide compatibility with model mixes based on NovelAI, which is most of them. Trained on a predominantly male dataset.