Search papers, labs, and topics across Lattice.
This paper introduces SAGA, a training-free method that enhances autoregressive video generation by addressing temporal errors such as flickering and motion jitter. By leveraging discrete latent acceleration as a spectral guidance signal, SAGA effectively stabilizes high-frequency perturbations without the need for retraining existing models. Experimental results show significant improvements in both temporal quality and image quality across multiple autoregressive diffusion models, indicating its practical applicability in high-quality video generation.
SAGA boosts temporal stability in autoregressive video generation, achieving a remarkable increase in temporal quality from 97.30 to 97.91 without retraining.
Autoregressive video diffusion enables efficient streaming and long-horizon video generation, but repeatedly reusing generated latents as causal context can amplify temporal errors, resulting in flickering, motion jitter, and structural drift. In this paper, we investigate this failure mode from a spectral kinematic perspective and identify discrete latent acceleration as an effective signal for revealing unstable high-frequency temporal perturbations. To this end, we propose SAGA, a training-free \textbf{\textit{s}}table \textbf{\textit{a}}cceleration \textbf{\textit{g}}uidance approach for \textbf{\textit{a}}utoregressive video generation. SAGA integrates an acceleration domain spectral guidance objective based on finite-window Slepian projections with a structured autoregressive noise initialization strategy that suppresses short-range temporal correlations while preserving long-range motion structure. Without retraining or modifying the backbone, SAGA can be directly applied to existing chunk-wise autoregressive diffusion models, which is the prevalent setting for high-quality generation. Extensive experiments show that SAGA consistently improves temporal quality across multiple autoregressive diffusion models. On Self-Forcing, SAGA improves Temporal Quality from 97.30 to 97.91 and Image Quality from 69.60 to 70.51. Moreover, spectral analysis and human preference studies demonstrate that SAGA reduces temporal instability while maintaining visual fidelity.
