What Is Real-Time AI Video?
Real-time AI video is generated or transformed quickly enough to participate in an ongoing viewing or interaction loop.
SHARED DEFINITIONS
“Real-time” can describe a completed clip, a continuous stream, or a world that reacts to actions. These definitions keep those products and their measurements separate.
CATEGORY DEFINITION
Real-time AI video is generated or transformed quickly enough to participate in an ongoing viewing or interaction loop.
BENCHMARK GLOSSARY
RTR is generation time divided by playable video duration. A value below 1.00 means generation completed faster than playback.
WORLD MODEL GLOSSARY
An interactive world model generates future observations conditioned on actions while maintaining enough state for a user or agent to continue interacting.
FROM TERMS TO DECISIONS
Use RTR for completed clips, first-frame latency and cadence for streams, and action-to-effect latency plus state persistence for interactive worlds.
Read the benchmark methodology