I trained the toki model with <a target="_blank" rel="ugc" href="https://civarchive.com/models/118729?modelVersionId=240824">Reproduction | SDXL - 2v1.2 | Stable Diffusion Checkpoint | Civitai</a>. I strongly recommend using it.And also I trained the lora with naive danbooru style tags.So, please prompt with tag style like toki (\bunny\) \(blue archive\), toki \(blue_archive\), halo, blond hair, long hair .....Sometimes, the model generates artifacts on ears because of earphones. So, earphones in negatimve prompts will help.I didn't test various strengths with my model but I think 0.6-1.0 should be good to generate.Enjoy ^^<hr /><a target="_blank" rel="ugc" href="https://civarchive.com/models/118729?modelVersionId=240824">Reproduction | SDXL - 2v1.2 | Stable Diffusion Checkpoint | Civitai</a>라는 SDXL로 학습 시켰습니다. 다른 모델은 잘 작동 안하더라고요.프롬프트 쓰실땐 단부루 태그 쓰듯이 쓰시면 됩니다. toki (\bunny\) \(blue archive\), toki \(blue_archive\), halo, blond hair, long hair .... 요런식으로요.바니걸 토키 생성할때, 이어폰이 이상하게 그려져서 이상하게 보일때가 종종 있는데, 네거티브에 earphones 넣으면 됩니다.로라 가중치는 0.6~1.0 정도가 적당한 거 같습니다.토키 많이 생성하셨으면 구독과 좋아요. 알림설정까지

Files with SHA256: edc67e245402e82e2ec8cff8ddde442ac8a6695b9701fc51800080af85319d8b