Moonvalley, an artificial intelligence (AI) startup based in Los Angeles, unveiled its generative video model called Marey on Wednesday. Named after the pioneering cinematographer Étienne Jules Marey, this innovative AI model is designed to operate exclusively on ethically sourced licensed data. The startup emphasizes that this careful procurement of training data minimizes legal risks, making it an attractive option for movie studios and filmmakers. Rather than presenting Marey as a standard text-to-video tool, Moonvalley aims to embed this AI technology within existing studio workflows.
Moonvalley Launches Marey AI Video Model
In a recent post on X (formerly Twitter), Moonvalley’s official account announced the introduction of its Marey video generation model, describing it as a “clean” technology tailored for filmmakers and built solely from licensed content.
The startup further elaborated in a blog entry that it seeks to transform the conventional “text in, video out” paradigm that limits filmmakers’ control over the creative process. The company noted that filmmaking is inherently iterative, and professionals desire tools that enhance their skills rather than simply deliver finished outputs.
The model diverges from typical video generation tools by providing various functionalities that allow filmmakers more authority over the final product, according to a report by TechCrunch. Marey purportedly offers customization features, including motion control and camera type options, as well as the ability to dictate the movements of specific elements within scenes. The report also highlighted that Marey is capable of generating HD video clips lasting up to 30 seconds.
Moreover, Marey will support multiple input formats, such as text, storyboards, sketches, photographs, and video clips. The company plans to deliver a robust software experience that integrates seamlessly with the tools and workflows that filmmakers currently employ.
Further specifics about the Marey AI model remain undisclosed at this time. Currently, a limited group of users has early access to the model, while others can join a waitlist for future opportunities to experiment with the technology.
