@seong_ryul

针对中景，远景进行了强化训练，特写会导致面部崩坏，不建议使用。另外水彩风格强烈，可能存在泛化性问题，酌情使用。同时采用了自然语言和tag混合抽样训练，自然语言表现较好，推荐使用。

This model has undergone specialized training for mid-range and long-range shots; however, close-ups may result in facial distortions and are therefore not recommended. Additionally, the watercolor aesthetic is quite pronounced, which may lead to generalization issues; please use with discretion. The model was trained using a hybrid sampling method combining natural language and tags; natural language prompts yielded superior results and are therefore recommended.