Join our daily and weekly newsletter for the latest updates and exclusive content on industry-leading AI coverage. learn more
A few years ago, there was no such thing as “generating AI video models”.
Today, there are dozens, including many that can present super high definition Hollywood-style videos within seconds of being able to prompt text or use images uploaded by users and clips of existing videos. If you have read VentureBeat in the past few months, there is no doubt that you will come across articles about these models and the companies behind them Gen-3 of the runway arrive Google’s VEO 2 Long-term delays for Openai, but ultimately available Sola arrive Luma ai,,,,, PickupAnd the new China Clint and Hailuo. even Alibaba and a startup called Genmo Open source video tape is provided.
These models have been used to create some major blockbusters Everything is everywhere at once arrive HBO’s Real Detective: The Night Country From music videos and TV commercials Toy r’we and Coca Cola. But while Hollywood and filmmakers embrace AI relatively fast, there is a huge looming problem: All rights reserved.
As far as we know, given that most AI video model startups do not publicly share the precise details of their training data, most people are trained in a large number of videos uploaded to the web or in a large number of videos collected from other archive sources, including those with copyright, including their owners who may or may not have given AI Video Companies training to train them. In fact, Runway is one of the companies facing class action lawsuits (still working in court) on this issue, and NVIDIA reportedly scratched a bunch of YouTube videos For this purpose, too. This dispute is underway about whether scratching data, including video, constitutes fair and changeable use.
But now, there is a new alternative for those who are concerned about copyright and don’t want to use a question mark model. A startup Moonvalley – Founded by researchers from Google Deepminders and Meta, Microsoft and Tiktok, and Marey, a generative AI videotape designed for Hollywood studios, film producers and corporate brands. Marey is positioned as a “clean” state-of-the-art basic AI video model, specially trained on owned and licensed data, providing an ethical alternative to AI models developed using scratched content.
“People say it is technically not feasible to build cutting-edge AI video models without using scratch data,” Moonvalley CEO and co-founder Naeem Talukdar said in a recent video call interview with VentureBeat. “We proved that.”
Marey, now available as an invitation list, join Adobe’s Firefly Video Modellong-established software vendors are also enterprise-level – only accepted licensed data and Adobe Stock Data (Shocking at Some Contributors) – And compensation for the use of the enterprise. Moonvalley is right Article 7 of this documentsaid it will defend its clients at its own expense.
Moonvalley hopes these features will make Marey appealing to large studios – even if Runway reached an agreement with them – and filmmakers, including countless new AI video creation options.
More “moral” AI videos?
Marey is the result of a collaboration between Moonvalley and Asteria, an artist-led AI film and animation studio. The model was built to assist, rather than replace creative professionals, providing filmmakers with new tools for AI-powered video production while maintaining traditional industry standards.
“Our belief is that unless you do this with the industry, you won’t get mainstream adoption in the industry,” Tarukda said. “The industry is largely clear that in order for them to actually use these models, we need to figure out how to build a clean model. Until today, the highest track is something you can’t do.”
Instead of scratching the internet for content, Moonvalley has established a direct relationship with the creator to license his footage. The company has spent several months building these partnerships and ensuring that all data for training is legally accessible and fully licensed.
Moonvalley’s licensing strategy is also designed to support content creators by compensating for their contributions.
“Now people are starting to hear what we’re doing, and most of our relationships are actually inbound,” Talukda said. “For the creators in the town, a lot of their videos are just sitting around. We want to help them make money and want to make an artist-centric model. Ultimately it’s a good relationship.”
Talukdar told VentureBeat that while the company is still evaluating and modifying its compensation model, creators are usually compensated based on the duration of their videos and pay them an hourly or minor fee under a fixed-term license agreement, such as 12 or four months. This allows potential recurring payments if the content continues.
The company’s goal is to make high-end video production more accessible and cost-effective, allowing filmmakers, studios and advertisers to explore AI-generated storytelling without legal or ethical issues.
More photography controls – beyond text prompts, images and camera orientation
Talukdar explained that Moonvalley takes a different approach to existing AI video models by focusing on professional-grade production rather than consumer applications.
“Most of the generated video companies today are more consumer-oriented,” he said. “They build simple models, here they motivate chatbots to generate some clips and add cool effects. Our focus is different: What technology does Hollywood studios need? What do major brands need to create Super Bowl ads?”
Marey introduces some advancements in AI-generated videos, including:
- Local HD generation – Generate high-definition video without relying on upgrades to reduce visual artifacts
- Extended video length – Unlike most AI video models that generate only a few seconds of footage, Marey can create 30-second sequences in a single pass.
- Layer-based editing – Unlike other generated video models, Marey allows users to edit the foreground, middle and background individually, thus more precisely controlling video composition.
- Storyboards and sketch-based inputs – Marey not only relies on text cues (like many AI models), but also enables filmmakers to create using storyboards, sketches and even live references, making it more intuitive to professionals.
- More sensitive to conditional input – The model is designed to better explain external inputs such as drawings and motion references, making AI-generated videos more controllable.
- “Generated” video editing – Moonvalley is developing companion software for Marey, a generated video editing tool that helps users manage projects and schedules more effectively.
“The model itself is just built around controllability,” Talukda explained. “You need to have more controls around the output – the ability to change characters. This is the first model that allows you to do layer-based editing, so you can edit the foreground, middle and background separately. It’s also the first specially made model for Hollywood to be used for production.”
Additionally, he told VentureBeat that Marey relies on a hybrid model of diffusion conversion that combines diffusion and transformer-based architectures.
“These models are diffusion-converter models, so are transformer architectures, and then you’ll diffusion as part of the layers,” Talukda said. “When you introduce controllability, it’s usually done through those levels.”
Funded by Big Name VC, but not as good as other AI video startups (not yet)
Moonvalley also announced this week the $70 million seed round led by Bessemer Venture Partners, Khosla Ventures and General Catalyst. Investors Hemant Taneja, Samir Kaul and Byron Deeter also joined the company’s board of directors.
Talukdar noted that Moonvalley has significantly less funding than some competitors, so far – the runway has reportedly Raised $270 million in several rounds – However, the company optimizes its resources by forming an elite team of AI researchers and engineers.
“We raised about $70 million, which is much less than our competitors,” he said. “But it does come down to the team – having a team that can build that architecture more efficiently, counting all these different things.”
Marey is currently in limited access, with selected studios and filmmakers testing the model. Moonvalley plans to gradually expand access in the coming weeks.
“Currently, there are a lot of studios visiting it, and we have an Alpha group where dozens of filmmakers are using it,” Talukdar confirmed. “Hopefully it will be fully used in a few weeks, the worst case scenario in a few months.”
With the launch of Marey, Moonvalley and Asteria aim to position themselves at the forefront of AI-assisted film production, offering studio and brand solutions that integrate AI without compromising creative integrity. But with competitors from AI video startups like runways, pickups and Hedra Continuing to add new features like character sounds and actions, the field is becoming increasingly competitive.
Source link