Updated: Dec 3, 2025
characterTrained with 78 images
If you like my work, please consider supporting me: https://ko-fi.com/suzushi2024
I am open to ideas and suggestions on characters lora, models etc.
Also built a free site to generate image and host models, more models are coming soon,
Try it out on : https://miyukiai.com/
The Zimg lora is experimental. Prompt can be a bit hard I am using xml to see how the model adapts to it.
The prompt style goes as the following.
<character></character>
<quality></quality>
<danbooru_tags></danbooru_tags>
<caption></caption>
Since the base model didn't really learn a lot of danbooru tags, the caption in natural language remains more important. But Z image generates anime style image pretty good already.




