Runway Unveils Real-Time Video Model Built with NVIDIA Hardware

Runway

Mar 20, 2026 · Updated Apr 25, 2026

Runway shared a research preview of a real-time video generation model developed with NVIDIA, running on Vera Rubin hardware. HD video generates instantly — time-to-first-frame under 100ms — opening a fundamentally different design space for video generation and world simulation.

Runway, the AI video generation company, shared a real-time video generation model as a research preview at NVIDIA GTC, co-developed with NVIDIA and running on Vera Rubin hardware. HD videos generate instantly, with time-to-first-frame under 100ms. Runway says this bolsters the foundations of GWM-1, their General World Model family.

Real-time generation opens a fundamentally different design space for video models and world simulation. Runway is investing in co-designing models alongside hardware advances to keep pushing this frontier — Gen-4.5 was ported from NVIDIA Hopper to Vera Rubin NVL72 in a single day, confirming backward compatibility across GPU generations.

This is a research preview with no public access announced yet. If you use Runway for video generation, watch for when real-time capabilities move from research into their production tools like Gen-4.5 and GWM-1.

View the full update on runwayml.com

Runway

@runwaymlMar 18

A breakthrough in real-time video generation. As a research preview developed with @NVIDIA and shared at @NVIDIAGTC this week, we trained a new real-time video model running on Vera Rubin. HD videos generate instantly, with time-to-first-frame under 100ms. Unlocking an entirely new creative paradigm and bolstering the foundations of our General World Model, GWM-1. Real-time generation opens a fundamentally different design space for video models and world simulation. We're investing in co-designing our models alongside advances in hardware to keep pushing this frontier.

257

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from Runway →

Keep reading

Runway Achieves 1.75 Second Latency for Real Time HD Video Agents

Runway shared technical benchmarks for its Characters API, achieving 1.75 seconds of end-to-end latency for conversational video agents. This performance shift moves AI video from a slow generation process to a live interactive interface capable of 24 frames per second in HD.

NVIDIAMay 22

NVIDIA Launches LongLive-2.0 for 4-Bit Long Video Generation Infrastructure

NVIDIA released LongLive-2.0, an end-to-end infrastructure that uses 4-bit floating point precision across the entire training and inference lifecycle for long-form video. By aligning 4-bit weights and activations with a multi-shot attention mechanism, the system achieves real-time speeds while maintaining subject consistency over minute-long generations.

Hao AI LabMay 27

Hao AI Lab Open Sources Dreamverse for Real Time Video Directing

Hao AI Lab released Dreamverse, an open-source reference application that generates 30-second 1080p videos in 7 seconds on a single NVIDIA B200 GPU. The system introduces vibe directing, a workflow where creators steer video generation through natural language in a real-time interactive loop.