Mar, 2024

视频电梯:用多功能的文本到图像扩散模型提升视频生成质量

TL;DRText-to-video diffusion models still lag behind compared to text-to-image models, so this paper introduces VideoElevator, a training-free and plug-and-play method that improves T2V performance using T2I capabilities, specifically by enhancing temporal consistency and adding photo-realistic details for personalized stylistic video synthesis.