该lora使用了纳米香蕉蒸馏出来的3d图片进行训练，同时加入了自然语言和tag混合训练，得到更好的画面遵循度。
This LoRa algorithm was trained using 3D images distilled from nano-bananas, and also incorporated natural language processing and tag-based training to achieve better image conformance.
Description
该lora使用了anima1.0正式版为base,使用纳米香蕉蒸馏出来的3d图片进行训练，同时加入了自然语言和tag混合训练，得到更好的画面遵循度，V2使用了更大的分辨率训练，理论上可以得到更好的细节，对训练集也进行了颜色矫正。
This LoRa dataset uses Anima 1.0 as its base and is trained using 3D images distilled from nano bananas. It also incorporates natural language processing and tag-based training to achieve better image fidelity. V2 uses a higher resolution for training, which theoretically can yield better detail. Color correction has also been applied to the training set.
FAQ
Comments (13)
很好的lora使我的迪克旋转
已经用上了，二弟觉得很棒
怎么有马😭😭
你牛大了
我的牛也大了
Wow, this is just amazing! Please keep updating. Thank you.
能不能出个光辉版的
感觉似乎有些过拟合？加了这个lora之后人物的动作几乎固定了
I cannot seem to replicate the image when i try the similar prompt , lora and checkpoint used. Im using stable diffusion forge neo
I absolutely love the way it looks. It's probably one of the only Lora models for Anima right now that works on its own, right out of the box.
I don't know why my character's posture not what I want in most of the time, maybe I should ture the cfg higher?
It is recommended to use detailed natural language descriptions.
效果非常棒！