Moonvalley Unveils Marey, a Licensed AI Video Model for Studios and Brands
As the number of generative AI video models continues to grow, so do the concerns surrounding the legality and transparency of the training data these systems rely on. While major players like Runway, Google, OpenAI, Pika, and others push the boundaries of AI-generated video content, questions around copyright and data sourcing remain unresolved.

As the number of generative AI video models continues to grow, so do the concerns surrounding the legality and transparency of the training data these systems rely on. While major players like Runway, Google, OpenAI, Pika, and others push the boundaries of AI-generated video content, questions around copyright and data sourcing remain unresolved. A new company, Moonvalley, is entering the conversation with Marey, an AI video model developed exclusively using fully licensed and owned content, aiming to provide a more transparent and legally sound alternative.
Founded by a team of former researchers from Google DeepMind, Meta, Microsoft, and TikTok, Moonvalley built Marey with the goal of offering a production-ready tool for professional studios, filmmakers, and enterprise brands. Instead of relying on scraped videos from the web, the company established direct licensing agreements with content creators, compensating them for the footage used in the model's training process. CEO Naeem Talukdar stated, “People said it wasn’t technically feasible to build a cutting-edge AI video model without using scraped data. We proved otherwise,” emphasizing the company's commitment to ethical development practices.
Marey is designed not for casual consumers, but for high-end use cases, including film production and large-scale advertising. The model supports native high-definition video generation, eliminating the need for upscaling. It can generate sequences of up to 30 seconds in a single pass—much longer than what most current models can produce. Moreover, it offers advanced controls like layer-based editing, enabling creators to manipulate foreground, midground, and background separately. This architecture gives filmmakers more precision and flexibility during the creative process.
One of the distinguishing features of Marey is its support for storyboard- and sketch-based inputs. Instead of relying solely on text prompts, users can feed the model visual inputs, including rough drawings or live-action references, to guide the generation process. “What’s the technology needed for Hollywood studios? What do major brands need to make Super Bowl commercials?” Talukdar asked, underlining Moonvalley’s focus on professional-grade production tools.
Technically, the model is based on a hybrid diffusion-transformer architecture. According to Talukdar, controllability is introduced through specific layers in the model, giving users detailed command over the generated content. Alongside Marey, Moonvalley is also developing a generative-native video editing platform to help users manage their creative workflow more efficiently.
In terms of funding, Moonvalley recently secured a $70 million seed round led by Bessemer Venture Partners, Khosla Ventures, and General Catalyst. Investors including Hemant Taneja, Samir Kaul, and Byron Deeter have also joined the board. While this amount is significantly lower than the funds raised by competitors like Runway, which has reportedly received $270 million, Moonvalley believes its lean and experienced team enables it to build powerful systems without excessive capital.
Currently, Marey is available through an invitation-only waitlist. Several studios and a select group of filmmakers are testing the model in its alpha phase. The company expects wider access to roll out in the coming weeks. Through its partnership with Asteria, an artist-led AI film and animation studio, Moonvalley aims to support creative professionals rather than replace them, ensuring that AI can be integrated into production workflows without compromising legal or artistic standards.