EVATok: Adaptive Length Video Tokenization for Efficient Visual Autoregressive Generation
This addresses a computational bottleneck in video generation for researchers and practitioners by improving efficiency without sacrificing quality, though it is incremental as it builds on existing tokenization methods.
The paper tackled the inefficiency of uniform token assignment in video tokenizers for autoregressive generation by introducing EVATok, which adaptively assigns tokens per video to optimize quality-cost trade-offs, resulting in at least 24.4% savings in average token usage and state-of-the-art performance on UCF-101.
Autoregressive (AR) video generative models rely on video tokenizers that compress pixels into discrete token sequences. The length of these token sequences is crucial for balancing reconstruction quality against downstream generation computational cost. Traditional video tokenizers apply a uniform token assignment across temporal blocks of different videos, often wasting tokens on simple, static, or repetitive segments while underserving dynamic or complex ones. To address this inefficiency, we introduce $\textbf{EVATok}$, a framework to produce $\textbf{E}$fficient $\textbf{V}$ideo $\textbf{A}$daptive $\textbf{Tok}$enizers. Our framework estimates optimal token assignments for each video to achieve the best quality-cost trade-off, develops lightweight routers for fast prediction of these optimal assignments, and trains adaptive tokenizers that encode videos based on the assignments predicted by routers. We demonstrate that EVATok delivers substantial improvements in efficiency and overall quality for video reconstruction and downstream AR generation. Enhanced by our advanced training recipe that integrates video semantic encoders, EVATok achieves superior reconstruction and state-of-the-art class-to-video generation on UCF-101, with at least 24.4% savings in average token usage compared to the prior state-of-the-art LARP and our fixed-length baseline.