StyleInV: A Temporal Style Modulated Inversion Network for Unconditional Video Generation

TL;DR

StyleInV introduces a novel inversion-based motion generator for unconditional video synthesis, enabling high-quality, temporally consistent, long-duration videos with less training complexity and support for style transfer.

Contribution

It proposes a learning-based inversion network for GANs that captures rich priors and modulates future latents, reducing reliance on heavy discriminators and enabling style transfer.

Findings

Outperforms existing methods in generating long, high-resolution videos

Achieves high temporal consistency and single-frame quality

Supports style transfer with simple fine-tuning

Abstract

Unconditional video generation is a challenging task that involves synthesizing high-quality videos that are both coherent and of extended duration. To address this challenge, researchers have used pretrained StyleGAN image generators for high-quality frame synthesis and focused on motion generator design. The motion generator is trained in an autoregressive manner using heavy 3D convolutional discriminators to ensure motion coherence during video generation. In this paper, we introduce a novel motion generator design that uses a learning-based inversion network for GAN. The encoder in our method captures rich and smooth priors from encoding images to latents, and given the latent of an initially generated frame as guidance, our method can generate smooth future latent by modulating the inversion encoder temporally. Our method enjoys the advantage of sparse training and naturally constrains the generation space of our motion generator with the inversion network guided by the initial frame, eliminating the need for heavy discriminators. Moreover, our method supports style transfer with simple fine-tuning when the encoder is paired with a pretrained StyleGAN generator. Extensive experiments conducted on various benchmarks demonstrate the superiority of our method in generating long and high-resolution videos with decent single-frame quality and temporal consistency.