The Hook
“In 2023, Runway introduced the first-ever commercially available video model, Gen-1. Ever since, our in-house research team has continued to push the frontier of what's possible with generative video models and how they're trained.”
Problem Agitation
“Now, we're taking those learnings and applying them to physical intelligence and robotics. Historically, every hour of robot training data had to be earned. Someone, somewhere, demonstrating one task at a time. Over, and over, and over again.”
Product Reveal
“With World Action Models, all videos of humans performing tasks become demonstrations, at a scale robot data cannot match. Introducing Praxis-1. An open-weights world action model built for any robot.”
Feature Teaser
“Praxis-1 learns how the physical world behaves from ordinary video. How objects move under contact. What a task looks like halfway through. And it does this all at scale. With a comparatively small amount of first-person and robot data, it can turn that understanding into control. Praxis-1 comes out of the same general world model research and real-time infrastructure running across Runway's Solaris and GWM Worlds 2. Those models learn the physical world by generating it. Praxis-1 uses that understanding to act within it. One model that can adapt to any embodiment with just a few minutes of demonstrations.”
Call to Action
“Praxis-1 open weights will be available soon. For a world that becomes more dynamic, alive, and editable.”
Synthesizable in Remotion · key components of the world-action-model-announcement template.
Centered media container with subtle rounded corners and clean sans-serif corner labels.
Wrap video in an absolute div with 40px padding, white background, and corner-positioned span labels using interpolate opacity.
Synchronized multi-camera array displaying camera angles with monospace task prompts.
Create a CSS grid of video tiles with a fixed header displaying current_task string typed out via frame counter.
Vertical ribbed semi-transparent diffusion layer with large center-aligned typography.
Apply backdrop-filter blur and repeating-linear-gradient lines over footage layer with Spring-animated text opacity.
Purpose. Establish credibility through generative video milestone timeline.
Execution. Rapid historical montage with minimal white borders.
Purpose. Highlight the inefficiency of physical robot teleoperation.
Execution. Multi-view repetitive arm capture clips with task labels.
Purpose. Introduce World Action Models learning directly from video.
Execution. Massive typography on fluted glass to human video mosaic.
Purpose. Demonstrate generalization across multiple physical tasks.
Execution. Robotic manipulation of delicate objects and spatial navigation.
Purpose. Announce open-weights release and provide URL.
Execution. Symmetrical architectural wide shots resolving to clear URL CTA.