Lora for Hitomi from doa. This was trained with 饭特稀_v0.8 which you can find on civit. The lora should work ok on similar models but it looks best on 饭特稀_v0.8. Trigger word is "hitomi".
Trained on a variety of her haistyles.
Also highly recommend EasyNegative which my sample images are using.
If you get bad looking faces at lower resolutions try inpainting only the face at higher resolution or using hires fix.
Description
FAQ
Comments (4)
thank you. I'm glad there are many character models.
Thank you! Hitomi is such a cute and powerful DoA character. The mentioned BaseModel 饭特稀_v0.8 and its newer versions can be found here:
https://civitai.com/models/18427/fantexiv09beta
ive tried so many versions and cant get good results using the example prompts
@gutzsy Hi. Didn't you get as good results as the ones I have posted below?
For the webui, I worked with: https://github.com/AUTOMATIC1111/stable-diffusion-webui
For the model, I worked with v.0.8 which you can download in the tab "..._v0.8" under the link I've posted in my frist comment. I got better results with a weight set higher than 0.7 in the prompt, e.g.:
<lora:doa_hitomi:0.9> ... ... ... or <lora:doa_hitomi:1.0> ... ... ...
Also set "Clip skip = 2" webui, under the 'Settings' tab > 'Stable Diffusion' on the left > scroll down to 'Clip skip'.
You have to try out. Also add more words to the Negative prompt to exclude what you don't like in your former result. You can see each of the prompts I made up by clicking on the (i) in the bottom right corner of my eight images below.
My first results were so-so, too. Use "Inpaint" to improve the generated image in a second step.
PLEASE, if you get showable results, post'em here! ;-) Good luck!