Meta, the parent company of Facebook and Instagram, revealed on Friday that they have developed a new artificial intelligence model named Movie Gen. This AI model can generate realistic video and audio clips based on user input, which can be easily used in major media outlets. Meta claims that Movie Gen is on par with similar tools from OpenAI and Eleven Lab.
Examples of Movie Gen’s output shared by Meta include videos of animals swimming and surfing, as well as images of people engaged in activities like painting. Movie Gen can also create background music and sound effects that synchronize with the video content, as mentioned in a blog post by Meta. Users can also utilize this model to edit existing videos.
Meta demonstrated the capabilities of Movie Gen by showcasing videos where a man inserts a pom-pom into another man’s hand while running in the desert, and turning a dry skateboarding area into a water-splashed parking lot. According to Meta, videos created with Movie Gen can be up to 16 seconds long, and the accompanying audio can last up to 45 seconds. Blind tests conducted by Meta show that Movie Gen performs favorably compared to similar products from various startups.
This announcement follows OpenAI’s presentation earlier this year of their product Sora, which can create feature-length videos resembling films based on text input. Technologists in the entertainment industry are excited about utilizing such tools to expedite film production, although concerns have been raised about potential copyright violations.
Concerns have also been raised by lawmakers about the use of AI-generated deepfakes in elections worldwide, including countries like the US, Pakistan, India, and Indonesia. Meta has stated that they are unlikely to release Movie Gen for open developer use, unlike their previous large-scale language models, and are carefully assessing the risks associated with each model individually.
Instead of a public release, Meta plans to collaborate directly with the entertainment industry and content creators to integrate Movie Gen into their products over the next year. According to Meta’s blog post and research paper, Movie Gen was developed using a combination of licensed and publicly available datasets.
OpenAI has been in discussions with Hollywood industry professionals regarding a possible partnership involving Sora, but as of now, no formal agreement has been reported. Concerns were raised about OpenAI when actress Scarlett Johansson accused ChatGPT developers of replicating her voice without permission for chatbots.
In a separate development, Lionsgate Entertainment, known for producing popular films like The Hunger Games and Twilight, announced a partnership with AI startup Runway to leverage their film and TV library for AI model training. This collaboration allows studios and filmmakers to enhance their creative work using the AI model.
Source: www.theguardian.com
