Runway Announces API for Its Video Generating Models

Runway, one of several AI startups developing video generation technology, today announced the availability of an API that will enable developers and organizations to embed the company’s generative AI models into third-party platforms, apps and services.

Currently in limited availability (there’s a waiting list), Runway API offers just one model to choose from—Gen-3 Alpha Turbo, a faster but less powerful version of Runway’s flagship Gen-3 Alpha—and two plans, Build (which is aimed at individuals and teams) and Enterprise. The base price is one cent per credit (1 second of video costs 5 credits), and Runway says that “trusted strategic partners,” including Omnicom’s marketing group, are already using the API.

The Runway API also has some unusual disclosure requirements. All APIs must “prominently display” a “Powered by Runway” banner linking to the Runway website, the company writes in a blog post, to “[help]users understand the technology behind (the apps) while complying with our terms of use.”

Runway, backed by investors including Salesforce, Google and Nvidia, and last valued at $1.5 billion, faces stiff competition in the video generation space, including from OpenAI, Google and Adobe. OpenAI is expected to release its video generation model, Sora, in some form this fall, while startups like Luma Labs continue to refine their technologies.

With the initial launch of the Runway API, Runway becomes one of the first AI vendors to offer a video generation model via API. However, while the API could help the company on its path to profitability (or at least recoup the high costs of training and running the models), it won’t resolve the unresolved legal issues surrounding these models and generative AI technology more broadly.

Runway’s video generation models, like all video generation models, were trained on a huge number of video examples to “learn” patterns in those videos and generate new content. Where did the training data come from? Runway declines to say, as do many vendors these days—partly out of fear of sacrificing competitive advantage.

But the training details also pose a potential source of intellectual property lawsuits if Runway trained on copyrighted data without permission. There’s evidence that’s exactly what happened — a July report from 404 Media revealed a Runway spreadsheet of training data that included links to YouTube channels belonging to Netflix, Disney, Rockstar Games, and creators like Linus Tech Tips and MKBHD.

It’s unclear whether Runway ultimately used any of the videos in the spreadsheet to train its video models. In an interview with TechCrunch in June, Runway co-founder Anastasis Germanidis would only say that the company uses “curated, internal datasets” to train its models. But if it did, it wouldn’t be the only AI vendor to play it safe with copyright.

Earlier this year, OpenAI CTO Mira Murati didn’t outright deny that Sora was trained on YouTube content. And Nvidia reportedly used YouTube videos to train its internal video generation model, called Cosmos.

Many generative AI vendors believe that a doctrine known as fair use provides legal protection — and they claim it in court and in public statements. Others are less willing to take risks and/or see a more “ethical” approach to model training as a selling point for their services. To expand his For example, Adobe is said to be offering artists payments in exchange for clips, via its Firefly video-generating models.

Regardless of how the lawsuits over the legality of copyrighted content training play out, one thing is becoming clear: Generative AI video tools threaten to upend the film and television industry as we know it. A 2024 study commissioned by the Animation Guild, the union representing Hollywood animators and cartoonists, found that 75% of film production companies that implemented AI reduced, consolidated, or eliminated jobs after incorporating the technology. The study also estimates that more than 100,000 entertainment jobs in the U.S. will be disrupted by generative AI by 2026.