Runway wants to beat Google

Runway said on May 15 that it is betting its next phase on video generation rather than treating video as a side effect of broader multimodal AI. The New York startup, which first gained traction with tools for filmmakers and editors, is now describing video as the route to what it calls “world models” — systems that can understand, simulate and act in environments. TechCrunch reported the strategy in an interview with co-founder and Chief Executive Cristóbal Valenzuela on Thursday. Runway has already aligned its product and fundraising around that thesis. The company said in February that it raised $315 million in Series E funding led by General Atlantic, and Valenzuela wrote at the time that the money would help “pre-train the next generation of world models and bring them to new products and industries.” Runway’s site now describes the company as “Building AI to Simulate the World.” (techcrunch.com) ### Why is Runway talking about video as more than a filmmaker tool? Runway’s current products are still rooted in media creation. Its Gen-4 page says the model is built for “media generation and world consistency,” and says users can generate consistent characters, locations and objects across scenes from reference images and text instructions. The company says that work improves control over motion, perspective and continuity — problems that matter in film production and in simulated environments. (runwayml.com) Cristóbal Valenzuela has tied that product work directly to a broader AI goal. In Runway’s February funding announcement, he called world models “the most transformative technology of our time,” while the company’s research page says it is building systems capable of understanding visual dynamics and acting in the world. ### What does Runway mean by “world models”? Runway’s homepage says its General World Models effort is aimed at systems that can “simulate all possible worlds and experiences.” The company’s GWM-1 product is described as a state-of-the-art general world model built to simulate reality in real time, with variants for explorable environments, conversational avatars and robotic manipulation. (runwayml.com) The company has been moving toward that language for months. (runwayml.com) TechCrunch reported on February 10 that Runway had released its first world model in December and was expanding beyond media and advertising into areas including gaming and robotics. ### Where does Google fit into this fight? Google DeepMind is already active in both video generation and world models. (runwayml.com) Its Veo page says Veo 3.1 is the company’s latest video generation model, with native audio and added controls for consistency and editing, while Google’s developer documentation says Veo 3.1 can generate 8-second videos at resolutions up to 4K. Google DeepMind has also published Genie 2 as a “large-scale foundation world model.” The company said Genie 2 can simulate virtual worlds and model object interactions, character animation and physics from large-scale video training. (techcrunch.com) ### Why does Runway think it can compete with a larger platform? Runway’s pitch to investors has emphasized specialization and execution. (deepmind.google) TechCrunch reported in February that the company’s latest funding followed the release of Gen 4.5 and a compute deal with CoreWeave, and said Runway had about 140 employees at the time with plans to expand research, engineering and go-to-market teams. (deepmind.google) Runway’s public materials also frame video quality and control as the near-term edge. Its homepage calls Gen-4.5 “the world’s best video model,” and the Gen-4 research page highlights consistency, physics and production workflows as core features rather than add-ons. Those claims sit alongside Google’s own push to fold video into broader Gemini and DeepMind products. ### What can users actually use now? (techcrunch.com) Runway’s website says Gen-4.5 is available now as its flagship video model, and it also lists GWM-1, Runway Characters and robotics-focused world-model tools among current offerings. The company says Characters is built on GWM-1 and lets users create conversational video agents from a single image without fine-tuning. Google’s current public lineup also appears to be live. (runwayml.com) DeepMind’s Veo page says users can try Veo in Gemini and Flow, and Google’s developer pages say Veo 3.1 is available through the Gemini API. Runway’s next visible milestone is product expansion tied to the February financing. The company said that funding would be used to pre-train new world models and bring them into additional industries, and its site is already advertising enterprise sales, Characters and robotics applications alongside Gen-4.5 and GWM-1. (runwayml.com 1) (runwayml.com 2) (deepmind.google)