AI Finally Reads the Rhythm of Motion

For the first time, AI systems developed by Google and University of Washington researchers can detect when video has been altered in time—from speeded-up clips to slow-motion footage—and generate motion at precise temporal rhythms. This matters because it fills a fundamental blind spot: machines could recognize objects and faces but couldn't perceive the subtle visual grammar of time, something humans instantly decode through motion blur and body dynamics. With this capability unlocked, AI can now power more sophisticated video forensics, detect deepfakes, and synthesize photorealistic motion—expanding the technology's reach far beyond today's applications.