Video Latent Flow Matching: Optimal Polynomial Projections for Video Interpolation and Extrapolation
This addresses video generation challenges for applications requiring flexible temporal control, though it builds incrementally on existing flow matching and pre-trained model techniques.
The paper tackles video generation by modeling a caption-guided flow of latent patches using pre-trained image models, achieving interpolation and extrapolation with arbitrary frame rates. Experimental results demonstrate effectiveness on multiple text-to-video datasets.
This paper considers an efficient video modeling process called Video Latent Flow Matching (VLFM). Unlike prior works, which randomly sampled latent patches for video generation, our method relies on current strong pre-trained image generation models, modeling a certain caption-guided flow of latent patches that can be decoded to time-dependent video frames. We first speculate multiple images of a video are differentiable with respect to time in some latent space. Based on this conjecture, we introduce the HiPPO framework to approximate the optimal projection for polynomials to generate the probability path. Our approach gains the theoretical benefits of the bounded universal approximation error and timescale robustness. Moreover, VLFM processes the interpolation and extrapolation abilities for video generation with arbitrary frame rates. We conduct experiments on several text-to-video datasets to showcase the effectiveness of our method.