Latent diffusion models pair VAEs with diffusion backbones, and the structure of VAE latents strongly influences the difficulty of diffusion training. However, existing video VAEs typically focus on reconstruction fidelity, overlooking latent structure. We present a statistical analysis of video VAE latent spaces and identify two spectral properties essential for diffusion training: a spatio-temporal frequency spectrum biased toward low frequencies, and a channel-wise eigenspectrum dominated by a few modes. To induce these properties, we propose two lightweight, backbone-agnostic regularizers: Local Correlation Regularization and Latent Masked Reconstruction. Experiments show that our Spectral-Structured VAE (SSVAE) achieves a 3× speedup in text-to-video generation convergence and a 10% gain in video reward, outperforming strong open-source VAEs. The code is available at https://github.com/zai-org/SSVAE.
We introduce Sparse Forcing, a training-and-inference paradigm for autoregressive video diffusion models that improves long-horizon generation quality while reducing decoding latency. Sparse Forcing is motivated by an empirical observation in autoregressive diffusion rollouts: attention concentrates on a persistent subset of salient visual blocks, forming an implicit spatiotemporal memory in the KV cache, and exhibits a locally structured block-sparse pattern within sliding windows. Building on this observation, we propose a trainable native sparsity mechanism that learns to compress, preserve, and update these persistent blocks while restricting computation within each local window to a dynamically selected local neighborhood. To make the approach practical at scale for both training and inference, we further propose Persistent Block-Sparse Attention (PBSA), an efficient GPU kernel that accelerates sparse attention and memory updates for low-latency, memory-efficient decoding. Experiments show that Sparse Forcing improves the VBench score by +0.26 over Self-Forcing on 5-second text-to-video generation while delivering a 1.11-1.17x decoding speedup and 42% lower peak KV-cache footprint. The gains are more pronounced on longer-horizon rollouts, delivering improved visual quality with +0.68 and +2.74 VBench improvements, and 1.22x and 1.27x speedups on 20-second and 1-minute generations, respectively.
Direct spectral editing in video-VAE latents can control noise, flicker, smoothness, and frequency content without a decode--filter--reencode pass. However, video VAEs may redistribute pixel-space frequency bands across latent channels, and latent edits can disrupt VAE round-trip dynamics. We introduce \emph{latent-frequency validity} (LFV), which learns a compact VAE-specific spectral response and deploys it only when it improves decoded-target fidelity without worsening round-trip drift. LFV follows a validation-selected path from a diagonal per-frequency calibrator (C1) to full channel mixing (CM), making cross-channel capacity a controllable per-edit resource. Across 544 VAE--edit cells spanning six spectral families, LFV emits 423 cheap operators: 277 are handled by C1, while 146 (34.5% of emitted operators) require channel mixing. On the primary 120-cell radial sweep, 99/100 emitted operators pass source-video-grouped held-out evaluation. Across five additional filter families, all 323 emitted operators pass held-out evaluation. Fully frozen OpenVid-fitted operators, including the validation-selected path coefficient, pass all 20 tested CogVideoX and HunyuanVideo generated-domain cells without adaptation. The selected response matches direct latent-filter latency and is about 3× faster than pixel filter--reencode. The resulting maps reveal distinct VAE regimes, including strongly channel-coupled CogVideoX responses and a sharp Open-Sora high-band stability frontier.
Video generative models commonly rely on latent spaces learned by 3D Variational Autoencoders (3D-VAEs). However, conventional 3D-VAEs are mainly optimized for pixel-level reconstruction, which can limit the semantic and spatio-temporal structure captured by their latents. Meanwhile, Video Foundation Models (VFMs) such as V-JEPA 2 and VideoMAEv2 show strong video understanding capabilities, yet whether their frozen representations can be transformed into compact, reconstruction-capable, and generation-friendly video latents remains largely unexplored. We answer this question with VideoRAE, a representation autoencoder that leverages multi-scale hierarchical features from a frozen video foundation encoder and compresses them with a lightweight 1D self-attention projector. VideoRAE supports both continuous latents for Diffusion Transformers and discrete tokens for autoregressive models via multi-codebook high-dimensional quantization. During decoding, a local-and-global representation alignment objective with the frozen VFM teacher improves semantic preservation and enables training without KL regularization. Experiments show that VideoRAE achieves strong reconstruction in both continuous and discrete regimes. On UCF-101, it obtains state-of-the-art class-to-video gFVDs of 40 and 93 with AR and DiT generators, respectively, while converging approximately 5x faster than competing autoencoder baselines. In a controlled 2B-scale text-to-video study, replacing LTX-VAE with VideoRAE leads to faster convergence under comparable settings. These results validate frozen VFM representations as versatile and generation-friendly video latents. The model and code will be released on https://zhxie0117.github.io/VideoRAE.