MotionGPT: A unified and versatile motion language model

Language instructions can be converted into 3D human movements!

For example, if you give the model a command, such as “kick” or “dance”, the model can generate corresponding 3D human movements.

The model converts 3D actions into action markers, similar to the process of generating word markers. Then treat human movements as a specific language, and train the actions and text in language modeling.

Hugging Face:https://huggingface.co/papers/2306.14795
Address: https://arxiv.org/pdf/2306.14795.pdf

Scroll to Top