Meta, the company led by Mark Zuckerberg, continues to advance in the field of generative artificial intelligence with the introduction of MovieGen, a family of models that allows the generation of videos with audio from text. This technology competes with options such as OpenAI's Sora and Gen-3 Alpha, offering powerful tools for the creation of audiovisual content.
Innovation in video and generative audio
MovieGen is divided into two main models. The first, MovieGen Video, is a 30 billion parameter transformer model capable of generating clips of up to 16 seconds at 16 frames per second, or 10 seconds at 25 frames per second, upscaled to 1080p resolution. On the other hand, MovieGen Audio, with 13 billion parameters, generates audio of up to 45 seconds that is synchronized with videos. Users can use prompts to adjust the results, such as adding ambient sounds or background music.
A standout feature of MovieGen is its ability to modify specific details of the generated videos without altering the overall content, making editing easier. This is useful for creators who need to make precise changes, such as altering the color of a garment or the direction of a movement.
Meta's AI research and future
Although Meta has shown impressive examples of MovieGen, such as a koala surfing or a person juggling fire, public access to this technology is not yet available. For now, MovieGen is still in the research phase in Meta's AI labs, but given the company's track record, it is likely to be available to users soon, following the path of other innovations, such as Meta AI chatbots on its social platforms.
With this new bet, Meta is positioning itself as a strong competitor in the creation of AI-generated audiovisual content, offering tools that could transform the way video is produced and edited in the future.
