![Face LoRA - kumiko_[Z-Image] by user_693186431694192228 on Tensor.Art](https://image.tensorartassets.com/cdn-cgi/image/anim=true,plain=false,w=500,q=85/model_showcase/980707806451600112/cba2307c-f8b6-8947-cb37-850ae449fc6e.png)
![Face LoRA - kumiko_[Z-Image] by user_693186431694192228 on Tensor.Art](https://image.tensorartassets.com/cdn-cgi/image/anim=true,plain=false,w=500,q=85/model_showcase/980707806451600112/108f90bc-e8ca-b165-d0a5-a3e4b3d84e5a.png)
![Face LoRA - kumiko_[Z-Image] by user_693186431694192228 on Tensor.Art](https://image.tensorartassets.com/cdn-cgi/image/anim=true,plain=false,w=500,q=85/model_showcase/980707806451600112/ce831a9b-be05-1712-6fbf-fcf526c0b9a0.png)
![Face LoRA - kumiko_[Z-Image] by user_693186431694192228 on Tensor.Art](https://image.tensorartassets.com/cdn-cgi/image/anim=true,plain=false,w=500,q=85/model_showcase/980707806451600112/cedfdb1f-d6bb-908b-f354-a297387197ce.png)
I created a set of three-view drawings with various expressions based on an AI-generated face and used them for training. As a result, it can output consistent faces from any angle and with any expression.
The optimal settings for this LoRA
Sampler : Euler
VAE : ae.sft (or None)
CFG scale : 1