✨ Multi-Character Model | MOSS Image

About This Model

This is a multi-character model created to test whether a dataset previously used to train multi-character models on other architectures can achieve similar results on MOSS Image.

So far, the results have been pretty good! ✨

With accurate and detailed prompts, each character maintains their distinctive features, with little to no feature bleeding between characters.

Prompting Tips

I recommend combining tags with natural language descriptions for better results. Of course, structured JSON prompts are even better for precise character control.

Character Generation

This model supports both single-character images featuring either of the two characters and two-character images featuring both characters together.

For detailed descriptions of each character's appearance, outfits, and distinctive features, please refer to the provided reference images.

----------中文说明----------

✨多角色模型 | MOSS Image

关于模型

这是一个多角色模型，主要用于测试此前在其他模型架构上训练多角色模型所使用的数据集，能否在 MOSS Image 上取得相近的效果。

目前的测试结果还不错！✨

在提示词足够准确、详细的情况下，各个角色的特征基本能够保持稳定，几乎不会出现不同角色之间的特征串扰或转移问题。

提示词建议

推荐使用 Tags + 自然语言描述 的组合，以获得更好的生成效果。当然，结构化 JSON 提示词通常更有利于精准控制角色特征。

角色生成

本模型支持生成两个角色各自的单人图片，也支持生成两个角色同时出场的双人图片。

关于角色的具体外貌、服饰及其他标志性特征，请参考我提供的角色参考图。