Research
Latent-Compressed Variational Autoencoder for Video Diffusion Models
arXiv:2604.16479v1 Announce Type: new Abstract: Video variational autoencoders (VAEs) used in latent diffusion models typically require a sufficiently large number of latent channels to ensure high-qu
arXiv:2604.16479v1 Announce Type: new Abstract: Video variational autoencoders (VAEs) used in latent diffusion models typically require a sufficiently large number of latent channels to ensure high-quality video reconstruction. However, recent studies have revealed that an excessive number of latent channels can impede the convergence of latent diffusion models and deteriorate their generative performance, even when reconstruction quality remains high. We propose a latent compression method that removes high-frequency components in video latent representations rather than directly reducing the number of channels, which often compromises reconstruction fidelity. Experimental results demonstrate that the proposed method achieves superior video reconstruction quality compared to strong baselines while maintaining the same overall compression ratio.
Related
- FrameDiT: Diffusion Transformer with Matrix Attention for Efficient Video Generation
- Latent Wavelet Diffusion For Ultra-High-Resolution Image Synthesis
- SVGFusion: A VAE-Diffusion Transformer for Vector Graphic Generation
- Measurement-Consistent Langevin Corrector for Stabilizing Latent Diffusion Inverse Problem Solvers
- HFI: A unified framework for training-free detection and implicit watermarking of latent diffusion model generated images
Source: arXiv cs.CV | 2026-04-21