Meta previewed new AI instruments on Friday referred to as Film Gen that may create movies, edit them mechanically, and layer on AI-generated sound for a cohesive video clip.
Film Gen works with written textual content prompts, a picture, or an present video as enter. There’s additionally an choice so as to add a private image so customers can see themselves within the video.
Associated: Meta Is Placing AI Pictures on Your Fb and Instagram Feeds, With Customized Photos
After the AI works its magic and generates a video, a consumer can kind in a textual content immediate to create a customized audio soundtrack to play with the video.
A peek right into a video Meta Film Gen created from a picture. Credit score: Meta
Whereas Meta says Film Gen’s high-definition movies are “the primary of its variety within the {industry},” when creating lengthy movies at totally different facet ratios, it doesn’t suggest the AI is ideal. Proper now the AI can solely generate movies that last as long as 16 seconds — and it does not at all times get the task proper.
In a demonstration to the New York Instances, Meta’s AI instrument made a mistake. Although it was capable of create a video of a canine in a park speaking right into a telephone, the AI tousled by inserting a human hand across the telephone as a substitute of a canine’s paw.
? At the moment we’re premiering Meta Film Gen: probably the most superior media basis fashions to-date.
Developed by AI analysis groups at Meta, Film Gen delivers state-of-the-art outcomes throughout a spread of capabilities. We’re excited for the potential of this line of analysis to usher in… pic.twitter.com/NDOnyKOOyq
— AI at Meta (@AIatMeta) October 4, 2024
Chris Cox, chief product officer at Meta, acknowledged in a Threads put up that Film Gen is “industry-leading” in video high quality however that Meta is not ready to launch the instruments as a result of they’re too costly and the movies presently take too lengthy to generate.
Meta is sharing what it has proper now as a result of the outputs “are getting fairly spectacular,” Cox wrote.
Meta is not the primary to indicate off a text-to-video AI generator instrument — ChatGPT-maker OpenAI did it in February with its text-to-video mannequin Sora.
In July, OpenAI printed a number of YouTube movies in partnership with artists and entrepreneurs exhibiting how Sora may create fantastical brief movies.
Associated: Mark Zuckerberg Does a Higher Job Than His Rivals at Explaining AI
