So, here’s to animators!
NVIDIA has developed a Text-2-Motion model — a tool for transforming text into moving images.
The main advantage is that it has open-source code and excellent documentation to help understand it.
What makes this tool special? Firstly, you can set constraints for the movements of different character parts — that is, limit their freedom, specify desired poses to achieve, define movement trajectories, and even control interactions with objects or use FK/IK systems.
The model was trained on over 700 hours of data with optical mockups — so it can do quite a lot.
Be sure to check this out:
https://research.nvidia.com/labs/sil/projects/kimodo/
There is a section about Humanoid Motion at NVIDIA, which discusses the company’s projects in Physical AI.
Also, this repository is worth exploring:
https://github.com/nv-tlabs/kimodo
Created with n8n:
https://cutt.ly/n8n
Created with syllaby:
https://cutt.ly/syllaby