latent-compression

2 posts tagged with "latent-compression"

July 9, 2026

Improved Video VAE for Latent Video Diffusion Model

ByteDance/Tongyi Lab's IV-VAE introduces Keyframe-based Temporal Compression and Group Causal Convolution to achieve SOTA video reconstruction with balanced inter-frame performance

video-vae latent-compression video-generation causal-convolution keyframe-compression

July 9, 2026

FSVideo: Fast Speed Video Diffusion Model in a Highly-Compressed Latent Space

ByteDance's FSVideo achieves 42.3x speedup over Wan2.1 via 64×64×4 compression autoencoder, layer memory self-attention DIT, and few-step latent upsampler

video-generation diffusion-model image-to-video latent-compression transformer