上次发布的 mo_da_03 其实训练的是 lokr，也是我做的第一个风格模型。当时的成品还有一些瑕疵，对原画风格的还原度也不够理想。经过这段时间的学习和进修，我决定重新训练一个 lora。所以，这次的模型才是真正的 lora，对原画风格的还原度相比之前也提升了不少。The previous version of mo_da_03 was actually trained as a lokr, and it was also the first style model I ever made. It had a few flaws, and its ability to reproduce the original art style was not quite where I wanted it to be.After spending some time learning and improving my training skills, I decided to retrain it as a lora. So this time, it’s an actual lora, with significantly better fidelity to the original art style than the previous version.

Files with SHA256: dc9a30d9d2bb2352c070ba776fd15e88fa42ed1951ed125f917e702a8d9ff7a1