Paper Overview
Research Area: Computer Vision (CV) Authors: Yen-Siang Wu, Rundong Luo, Jingsen Zhu, Tao Tu, Ali Farhadi, Matthew Wallingford, Yu-Chiang Frank Wang, Steve Marschner, Wei-Chiu Ma Published: 2026-04-23 arXiv: 2604.21931
Abstract
How can we tell whether a video has been sped up or slowed down? How can we generate videos at different speeds? Although videos have been central to modern computer vision research, little attention has been paid to perceiving and controlling the passage of time. In this paper, we study time as a learnable visual concept and develop models for reasoning about and manipulating the flow of time in videos.
Key Contributions
- Temporal reasoning via self-supervision: The authors exploit multimodal cues and temporal structure naturally present in videos to learn, in a self-supervised manner, to detect speed changes and estimate playback speed.
- Largest slow-motion dataset: These learned temporal reasoning models enable curation of the largest slow-motion video dataset to date from noisy in-the-wild sources. Such slow-motion footage, typically filmed by high-speed cameras, contains richer temporal detail than ordinary video.
- Temporally controllable models: Using this data, the authors develop models for:
- Speed-conditioned video generation — producing motion at a specified playback speed.
- Temporal super-resolution — converting low-frame-rate, blurry videos into high-frame-rate sequences with fine-grained temporal detail.
Significance
The findings highlight that time is a manipulable perceptual dimension in video learning, opening doors toward temporally controllable video generation, temporal forensics (detecting manipulated footage), and potentially richer world models that understand how events unfold over time.
---
*Auto-collected on 2026-04-25 from zhichai.net.*