This is a lora of a neural adversarial network. When censorship is necessary, you can add this lora and use it simply. Because I found that when using the wan2.1 model in some networks, it will cause output content censorship. This is an amazing discovery! The weight defaults to 1. If the censorship is too heavy, you can use it from 3 and the gradient downgraded to 1 to reasonably resist neural network censorship.