Type | |
Stats | 231 |
Reviews | (13) |
Published | Jun 22, 2023 |
Base Model | |
Training | Steps: 3,500 |
Trigger Words | Sj0kz123 |
Hash | AutoV2 E2B48539A1 |
also made an embedding from the same data if you want a smaller file but you can use both at the same time really, set the lyco strength to like 0.3-0.6 if you're gonna do that: https://civitai.com/models/99013/eefje-depoortere-sjokz-textual-inversion
this is my first attempt at any sort of model and i have no clue what im doing
using a strength of 0.8-0.9 seems to work, maybe i overtrained, idk
gens seem to be mostly closeups cause thats the vast majority of my training, use controlnet or prompt better than i can for anything further out
example images used dpm++ 2m k, CFG7, 512x768, no fixing or upscaling or anything else, using deliberate and reliberate, and some prompts that i basically stole and forgot where from so sorry if it was you
DONT BE FKIN WEIRD OKAY
35 images, 100 steps each, 512x512, v1.5 base model
this is a loha, not a lora, so you'll need the lycoris extension: https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris