Efficient Video Prediction via Sparsely Conditioned Flow Matching

TL;DR

The paper presents RIVER, a novel efficient video prediction model that uses sparse conditioning, latent space modeling, and flow matching to generate high-quality videos with reduced computational costs.

Contribution

RIVER introduces a sparsely conditioned flow matching approach in latent space, enabling faster and more resource-efficient high-resolution video prediction.

Findings

Achieves comparable or superior performance to prior models.

Requires significantly fewer computational resources.

Enables high-resolution video generation with faster inference.

Abstract

We introduce a novel generative model for video prediction based on latent flow matching, an efficient alternative to diffusion-based models. In contrast to prior work, we keep the high costs of modeling the past during training and inference at bay by conditioning only on a small random set of past frames at each integration step of the image generation process. Moreover, to enable the generation of high-resolution videos and to speed up the training, we work in the latent space of a pretrained VQGAN. Finally, we propose to approximate the initial condition of the flow ODE with the previous noisy frame. This allows to reduce the number of integration steps and hence, speed up the sampling at inference time. We call our model Random frame conditioned flow Integration for VidEo pRediction, or, in short, RIVER. We show that RIVER achieves superior or on par performance compared to prior work on common video prediction benchmarks, while requiring an order of magnitude fewer computational resources.