April 1, 2012 - Artificial intelligence (AI) startups Runway on Monday released its newly developed AI Video Generation Model Gen-4. The company claims thatThis is one of the highest fidelity AI-driven video generation tools to date.
The newly released Gen-4 model is now being rolled out to Runway's individual and enterprise customers. According to RunwayThe core strength of the model is the ability to maintain a high degree of consistency of characters, locations and objects across video scenes, maintaining a "coherent world environment".and the ability to regenerate elements from different viewpoints and locations within the scene.
In its official blog post, Runway states, "Gen-4 is able to effectively utilize visual references and combine them with textual commands from the user to create new images and videos that are consistent in style, subject, location, and other elements, with no model fine-tuning or additional specialized training required throughout the process."
Backed by leading investment organizations including Salesforce, Google, and NVIDIA, Runway is focused on providing a range of AI video creation tools including Gen-4. However, Runway faces strong challenges from tech giants such as OpenAI and Google in the highly competitive arena of AI video generation. In order to differentiate itself in the market, Runway has adopted a differentiated strategy, not only by entering into a partnership agreement with a major Hollywood studio, but also by allocating millions of dollars to fund film projects that utilize AI-generated video technology in their creation.
In terms of Gen-4's functionality, Runway says that users simply provide a reference image of the character, and the model generates characters that consistently look the same under different lighting conditions. When building a specific scene, users can upload an image of the subject, accompanied by a textual description that specifies the compositional requirements of the shots they wish to generate.
In a blog post, Runway further emphasized that "Gen-4 excels at generating videos with a high sense of dynamism and realistic motion effects, while achieving top-of-the-class levels of subject, object and style consistency, precise adherence to user commands, and understanding of real-world laws." The company also declared, "The release of Runway Gen-4 also marks a significant milestone in the visual generation model's ability to simulate real-world physical laws."
Like all current video generation models, Gen-4 was trained on a huge amount of video data. By learning patterns from this data, the model is able to generate new synthetic video clips. However, Runway declined to disclose the exact source of its training data, partly to protect its commercial competitive advantage, but also because the details of the training data are often a sensitive area for potential intellectual property litigation.
As a prime example, Runway is currently facing a lawsuit filed by a group of artists. The lawsuit alleges that Runway and several other generative AI companies have used copyrighted artwork to train their AI models without authorization. In response, Runway has argued that its actions are protected from liability under the Fair Use doctrine. The final outcome of the case is unclear.
The stakes are high for Runway in this legal battle. According to The Information, Runway is currently in the midst of a new round of funding that could value the company at $4 billion (note: current exchange rate is about 29.044 billion yuan). With new products such as its Video Generation Model API (Application Programming Interface), Runway expects to reach its goal of $300 million in annualized revenue this year.
Regardless of where this lawsuit goes, generative AI video tools, represented by Gen-4, are posing a disruptive challenge to the existing film and television industry. A 2024 study commissioned by the Animation Guild, which represents Hollywood animators and cartoonists, revealed that up to 75% of film production companies that have already embraced AI technology have seen layoffs, consolidation of positions, or outright elimination of positions after introducing the technology. The study also predicts that by 2026, generative AI could impact more than 100,000 U.S. entertainment industry jobs.
