这个模型仅仅是我自己训练出来给我的国漫角色lora生图用的，缺点很严重很明显，出图建议半身图（upper body）或者脸图 (portrait)，配合 lora https://civarchive.com/models/1355454/25d-artmixstyle?modelVersionId=1599233 使用

This model was trained solely by me for generating images with my Chinese anime-style character LoRA. It has significant and obvious drawbacks, so it's recommended to generate upper-body or portrait images when using it, in combination with the LoRA at https://civarchive.com/models/1355454/25d-artmixstyle?modelVersionId=1599233

为什么会发布一个 WIP 版本，主要是因为收集素材太费时间了，这个版本是我通过 1200 张图片素材训练出来的（人像 800 张，背景 400 张），可想而知如果收集到 5000 张素材需要多久

The reason for releasing a WIP version is mainly that collecting training materials takes too much time—this version was trained on 1,200 images (800 portraits and 400 backgrounds), so you can imagine how long it would take to gather 5,000 images.