The conventional merge model may be contaminated with NAIleak models or DreamlikeLicense,
There were some situations where it was difficult to make decisions during full-scale business use.
This time, BracingEvoMix talked directly with Mr. BanKai, the developer of BRA, and
This is a merged model based on OpenBraβ and OpenBra, which were created as a result of donations from volunteers.
All the others are learning models, which greatly reduces the probability of risk model contamination compared to combining them with a merge model.
It is considered to be the lowest risk model among the merged models that feature Far East Asian faces.
However, even if it is called a learning model, it is not risk-free because we cannot know the details of its content.
We also consider factors such as cosine matching rate, but please be aware that we cannot guarantee that there will be no contamination.