v3: Three versions, trained for 6000 steps using ST based on 18 images of 1024x1024 resolution. Essentially resolves the issue of blurriness and can generate characters in non-anime styles.
When generating characters in non-anime styles, it is necessary to appropriately reduce the Lora weights to avoid overfitting, which can lead to disordered lines.
v3: Three versions, trained for 6000 steps using ST based on 18 images of 1024x1024 resolution.
Essentially resolves the issue of blurriness and can generate characters in non-anime styles.
Trigger word: (girl Verina/Verina)
Trigger word (anime style): (anime girl Verina)
When generating characters in non-anime styles, it is necessary to appropriately reduce the Lora weights to avoid overfitting, which can lead to disordered lines.
v3：三个版本，使用st基于18张1024x1024的图片进行训练了6000步。
基本上解决了模糊问题，而且可以生成非动漫风格的人物。
触发词：（girl Verina/Verina）
触发词（动漫画风）：（anime girl Verina)
生成非动漫风格的人物时，需要适当降低Lora权重以避免出现过拟合导致的线条紊乱。