Meta, the parent company of Facebook and Instagram, has just taken a massive leap into the AI sphere with the introduction of its newest model: Movie Gen. The AI-powered tool is capable of producing realistic video and audio clips based on user prompts, placing it as a rival to other media generation platforms such as OpenAI’s Sora and ElevenLabs.
In a press statement, Meta announced a few examples generated by Movie Gen, which range from mesmerizing visuals of animals swimming to surfing footage and videos where real photos morph into action shots. Movie Gen’s most remarkable aspect is its ability to create background music and sound effects appropriate to the videos’ content. Interactivity and personalization at this level set Movie Gen apart in the cutthroat world of AI-powered media. Videos created with Movie Gen can last as long as 16 seconds, while the audio can last up to 45 seconds. Blind tests showed promising results, placing the model well against the products of competitors like Runway and Kling, which means this model is already viable for wide-scale usage.
Yet the launch of Movie Gen arrives when Hollywood is still working out how to incorporate generative AI technologies into the filmmaking process fully. In January, for example, OpenAI’s Sora was able to generate feature-film quality video from rudimentary text input—a development that delighted technologists looking at various ways to deploy such systems as ways to accelerate the filmmaking process.
Meta’s decision to enter this sphere comes with substantial competition and opposition. OpenAI, Meta’s most significant competitor in the sphere, has also met with Hollywood executives and agents to discuss potential deals related to Sora. However, contracts have yet to be signed to allow a deal like this to proceed. Meanwhile, Lions Gate Entertainment, the studio behind blockbuster franchises like “The Hunger Games” and “Twilight” has recently partnered with another AI startup, Runway. This partnership gives filmmakers access to its extensive film and TV library and allows them to leverage AI models to enhance creative processes.
Concerns over copyrights and ethics dampen enthusiasm for tools of this sort as the AI models have to be trained with old content, often without forthright permission. Alarm bells have gone off in the legislatures over the potential for misuse of AI-generated content, especially in politics, as deep fakes capable of possibly manipulating election outcomes.
Spokespeople for Meta say it won’t release Movie Gen for open use by developers, as it has done with its Llama series of large-language models. Instead, the company has focused on direct collaborations with the entertainment community and other content creators, intending to incorporate Movie Gen into Meta’s products by next year.
It could be that as the landscape of video generation in AI continues to evolve and find its place in the mainstream, Movie Gen might usher in a new frontier in the way that content is created and digested. This may lead to new frontiers that had been taken for granted and reshaping the future in ways not previously imagined.
