Back to All Resources

LVDM (Latent Video Diffusion Model)

Developed by Tencent AI Lab, LVDM is a latent video diffusion model designed for high-fidelity long video generation. By leveraging a low-dimensional 3D latent space, LVDM efficiently generates videos with thousands of frames. Its hierarchical diffusion approach enables the production of extended video sequences, while addressing common errors in long video generation through techniques like conditional latent perturbation and unconditional guidance. LVDM has demonstrated state-of-the-art performance on various benchmarks, making it a significant advancement in AI-driven video generation.