Purpose
This model attempts to achieve real hip grinding that is difficult to prompt for. This is an early test model, so expect issues.
This has specifically been trained for i2v.
Usage
Motion seems to be best with strength around 1.0 - 1.5. Trigger phrase is "her hips grinding".
Known Issues
Not trained on back view.
Tends to make people sing if your scene has music and the strength is too high. My best guess as to why this is happening is because grinding hip movement looks a lot like singing / dancing hip movement that is part of the base H3 model. I likely just need a larger training dataset to counteract this.