Abstract
Contemporary models for generating images show remarkable quality and versatility. Swayed by these advantages, the research community repurposes them to generate videos. Since video content is highly redundant, we argue that naively bringing advances of image models to the video generation domain reduces motion fidelity, visual quality and impairs scalability. In this work, we build Snap Video, a video-first model that systematically addresses these challenges. To do that, we first extend the EDM framework to take into account spatially and temporally redundant pixels and naturally support video generation. Second, we show that a U-Net - a workhorse behind image generation - scales poorly when generating videos, requiring significant computational overhead. Hence, we propose a new transformer-based architecture that trains 3.31 times faster than U-Nets (and is ~4.5 faster at inference). This allows us to efficiently train a text-to-video model with billions of parameters for the first time, reach state-of-the-art results on a number of benchmarks, and generate videos with substantially higher quality, temporal consistency, and motion complexity. The user studies showed that our model was favored by a large margin over the most recent methods. See our website at this https URL.
Abstract (translated)
当代生成图像的模型具有出色的质量和多样性。受到这些优势的启发，研究社区将它们用于生成视频。由于视频内容高度冗余，我们认为过于简单地将图像模型的进步应用到视频生成领域会降低运动质量、视觉质量和可扩展性。在这项工作中，我们构建了Snap Video，一种视频优先的模型，系统地解决这些挑战。为此，我们首先将EDM框架扩展到考虑空间和时间冗余的像素，并自然支持视频生成。然后，我们证明了当生成视频时，U-Net - 负责图像生成的关键技术 - 表现不佳，需要大量的计算开销。因此，我们提出了一个基于Transformer的新架构，该架构训练速度比U-Nets快3.31倍（并且在推理时约快4.5倍）。这使得我们能够高效地训练具有数十亿参数的文本到视频模型，达到一些基准测试的最好结果，并生成具有相当高的质量、时间和运动复杂性的视频。用户研究表明，我们的模型在最近的方法中优势明显。请查看我们的网站：https://www.thisurl.com。
URL
https://arxiv.org/abs/2402.14797
