Type | |
Stats | 2,047 |
Reviews | (306) |
Published | Mar 4, 2023 |
Base Model | |
Training | Steps: 5 Epochs: 25 |
Trigger Words | symmetra blue dress visor thighhighs one mechanical arm |
Hash | AutoV2 BF81B2D7B6 |
My seventh Overwatch LORA. Focused on original appearance.
Triggers - symmetra, blue dress, one mechanical arm, visor, thighhighs
Headgear - Head acessories besides blue visor tint are glued and you have to really try to take them off , if you add cyborg, earrings, forehead jewel, headgear, jewelry, necklace, visor into negatives you might just get lucky and get rid of most of it.
Checked strength range is 0.6-0.9. Inpainting eyes works better on lower strenghts, face - on higher.
This one was pain in the ass to make, goddamn visors and headgears + homogeneous dataset = suffering.
If you liked my work, you can donate here.
Review and comment with your experiences, I would appreciate whatever feedback you can provide in order to continue improving my technique.
Trained on Anything V4.5 with anime.vae.pt.
Samples are 512x768, without HighRes.fix (mostly), using OrangeCocoa 50/50 Mix , with OrangeMix VAE, and using EasyNegative embedding, CLIP Skip 2. Inpaint examples are txt2img with 1 single inpaint done on face (I will comment on exceptions).