Los Angeles-based startup Moonvalley has launched an AI video generation model.
Named “Marey” after the film “Trailblazerétienne-Jules Marey”, the model was built in collaboration with Asteria, a new AI animation studio. According to Moonvalley, Marey is trained in “owned or fully licensed” source data and offers customization options such as fine-grained cameras and motion control.
“Marie gives you subtle control over movements within the scene,” Moonvalley wrote in a press release provided to TechCrunch:
The wide availability of tools for building video generators has resulted in a Cambrian explosion of vendors within the space. In fact, it risks oversaturation. Startups like Runway and Luma have released models with high-speed clips by tech giants like Openai and Google. In many cases, they are rarely distinguished from each other.
Moonvalley is selling Marey. From a legal perspective, Marey can generate “HD” clips up to 30 seconds, which are less risky than their competitors.
Moonbury is go! 🌗🚀
As many people know, I’ve done a lot in the field of video and animation over the last few months, and seeing this model built behind the scenes has been thrilling!
I was delighted to have had the opportunity to start playing with Marey, the world’s first 100%…pic.twitter.com/ddl4kwehrt
– Araminta (@aramineta_k) March 12, 2025
Many generated video startups train models of public data, some of which are always copyrighted. These companies argue that fair doctrine protects practices. However, it does not stop the rights owner from filing a complaint, halting or misleading.
Moonvalley says they are working with partners to handle license arrangements, packaging the videos and packaging them into data sets that the company purchases. This approach is similar to Adobe’s, which procures video footage of training from creators through Adobe’s inventory platform.
Many artists and creators are wary of video generators, and of course they are threatening to overturn the film and television industry. A 2024 study commissioned by the Animation Guild, a union representing Hollywood animators and cartoonists, estimates that by 2026 more than 100,000 US-based film, television and animation jobs will be destroyed by AI.
Moonvalley plans to require creators to remove content from the model, allowing customers to delete data at any time, and provide compensation policies to protect users from copyright issues.
Unlike some “unfiltered” video models that easily insert people’s portraits into clips, Moonvalley also promises to build guardrails around creative tools. Like Openai’s SORA, Moonvalley’s model cannot block certain content, such as NSFW phrases, and encourage people to generate videos of certain people or celebrities.
“It proves that it is possible to train AI models without bravely stealing creative work from cinematographers, visual artists, creators and creative producers,” said Talukdar, co-founder and CEO of Moonvalley in a statement. “As this technology and industry evolve, Moonvalley sets new standards for generation AI to provide industry-leading AI capabilities while ensuring that creative voices and rights are not lost.”