Runway

Three artists walked into NYU's ITP program in 2015 and came out with the world's most important video AI company. That's not a metaphor. Cristóbal Valenzuela, Alejandro Matamala, and Anastasis Germanidis were not ML researchers chasing benchmarks. They were makers — one part engineer, one part designer, one part artist — who believed that the tools for visual storytelling were embarrassingly primitive relative to what technology could enable. So they built better ones.

Seven years and $1.05 billion in funding later, Runway's Gen-4.5 is the top-rated video generation model in the world. Its technology was used in the production of Everything Everywhere All at Once. Lionsgate has trained a custom Runway model on 20,000+ films from its IP library. NVIDIA and Google are both investors. And Runway just launched GWM-1 — its General World Model — pointing toward a future where AI doesn't just generate video but simulates the physics and causality of entire worlds.

The Model Family, Explained

Runway ships and iterates faster than any other frontier video AI lab. Here is the current state of the lineup:

• Gen-4.5 — the flagship text-to-video and image-to-video model; top-rated for visual fidelity, cinematic motion, character consistency across scenes, and camera control; generates up to 16 seconds of video per clip
• Gen-4 Turbo — faster, cheaper generation for iteration and exploration; 5x the generation speed of Gen-4 at fraction of the credit cost
• Act-Two — full-body performance capture that animates any character from a driving video of a human performer; no motion capture suit, no studio required
• Aleph — in-video editor that applies generative changes (style transfer, object replacement, scene transformation) to existing footage while preserving motion, structure, and timing
• Veo 3.1 and Veo 3 — Google's video generation models available natively inside the Runway workspace via partnership
• GWM-1 — General World Model; Runway's research frontier; generates not just video but interactive, navigable simulations of physical environments — the foundation of what Runway calls "the next medium"

What $12/Month Actually Gets You

The Standard plan at $12/user/month is one of the most generously featured entry tiers in AI video. 625 credits per month — enough for 25 seconds of Gen-4.5, or 125 seconds of Gen-4 Turbo — with access to every model in the lineup including Veo 3.1 and Act-Two. Watermark removal. Resolution upscaling. Unlimited video editor projects. 100GB of asset storage. For a filmmaker, creative director, or content studio comparing options, there is no serious competitor at this price point with this model access.

The Unlimited plan at $76/month removes the generation ceiling entirely for Explore Mode — letting creators iterate freely on Gen-4.5, Act-Two, and Gen-3 Alpha Turbo without watching a credit counter. For production teams and agencies with high-volume needs, it is the more economical choice at scale.

The Enterprise Partnerships That Validate the Vision

• Lionsgate — custom AI model trained on 20,000+ films and TV titles from the Lionsgate library; enables IP-safe generative production assets
• NVIDIA — strategic investor and compute partner; Runway's video models run on NVIDIA's Vera Rubin architecture
• Google — strategic investor; Veo 3 and Veo 3.1 natively integrated into the Runway workspace
• AMC Networks — AI-powered content creation and marketing campaigns
• KPF Architecture — using Runway to animate architectural renderings and accelerate the visualization pipeline
• UCLA Film, Television and Digital Media — curriculum integration for the next generation of filmmakers

Runway Labs: The Incubator

In early 2025, Runway launched Runway Labs — a generative AI incubator led by co-founder and Chief Innovation Officer Alejandro Matamala Ortiz. Labs is exploring applications of video generation and world models across healthcare, education, gaming, real estate, and advertising — partnering with creators, enterprises, and institutions to find use cases beyond entertainment. It is Runway's explicit bet that the General World Model is not a video tool. It is a new medium entirely.