Trained 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.