Kyutai Unveils CADRE Video Autoencoders to Reduce World Model Lag
Research lab Kyutai has released CADRE, a family of seven open causal diffusion video autoencoders designed for interactive world models. Operating at 20 frames per second on a single H100 GPU, the technology achieves a reaction time of under 230 milliseconds. This breakthrough significantly improves real-time processing capabilities for generative video and interactive artificial intelligence systems.
Source: MarkTechPost