What happened
Researchers built a text-to-video AI model that generates video continuously as you watch, rather than finishing an entire clip before you see anything. This means interactive applications like live video editing, streaming generation, and real-time video effects become possible instead of requiring a 30-second wait for rendering.
Why it matters
Video generation has been stuck in a batch-processing trap — you request a video, wait minutes while a computer renders it offline, then watch the result. This removes that bottleneck by making generation happen live, which structurally changes what kinds of applications are even possible to build.