Search papers, labs, and topics across Lattice.
1
0
3
Achieve versatile and controllable video generation and editing using diffusion transformers, without any video training data, by cleverly reparameterizing the architecture and training only on 2D images.
