Runway and NVIDIA show off sub-100ms real-time video

Waiting several minutes for an AI video to render might soon be a thing of the past.

At this week's NVIDIA GTC, Runway and NVIDIA dropped a research preview of a new real-time video generation model that turns waiting time into blink-and-you-miss-it speed.

Instant HD generation at sub-100ms

Running on NVIDIA's Vera Rubin hardware architecture, the model generates high-definition video instantly, pushing the time-to-first-frame latency under 100 milliseconds.

Runway noted that this milestone is the result of co-designing models directly with hardware advances. The company plans to use this breakthrough to reinforce the foundations of its General World Model, GWM-1.

For now, the tech remains an early research preview, with no public release date or pricing details shared yet.

Why it matters

AI video is shifting from offline rendering to real-time simulation. When latency drops below 100ms, AI generation stops being a slow creative waiting game and starts looking like an interactive rendering engine suitable for gaming, real-time virtual production, and interactive worlds.

We think Runway’s move to tie its algorithm directly to NVIDIA’s next-gen silicon is a smart play to own the interactive frontier early. That said, the catch is hardware dependency: as long as this performance relies on cutting-edge Rubin chips, it will take some time before everyday creators can tap into it at scale and on a reasonable budget.