Search papers, labs, and topics across Lattice.
1
0
2
7
VideoRAE reveals that frozen Video Foundation Model representations can be effectively transformed into high-quality, generation-friendly video latents, achieving state-of-the-art results with faster convergence.