- Meta simply launched Movie Gen, an AI video generator to compete with OpenAI’s Sora.
- Movie Gen could make movies with accompanying audio utilizing a textual content immediate. It can even edit by immediate.
Meta launched a brand new AI video-generating instrument on Friday that can also be the corporate’s newest volley in its battle with OpenAI for AI supremacy.
“Today, we’re excited to premiere Meta Movie Gen, our breakthrough generative AI analysis for media, which incorporates modalities like picture, video, and audio,” the corporate mentioned in a press launch. “Movie Gen outperforms comparable fashions within the trade throughout these duties when evaluated by people.”
In its press launch, Meta referred to as Movie Gen the “most superior and immersive storytelling suite of fashions,” together with video technology, audio technology, customized video technology, and video modifying. The fashions have been educated utilizing publicly accessible knowledge and licensed knowledge, the corporate mentioned.
With a textual content immediate, Movie Gen can create movies as much as 16 seconds lengthy at 16 frames a second whereas reasoning “about object movement, subject-object interactions, and digicam movement.” Users can add an image of themselves to be included into customized movies, and Movie Gen can edit movies with textual content directions from the person.
Meta’s instance video reveals an underwater perspective of a child hippo (Moo Deng reference, anybody?) fortunately swimming round in a serene aquatic scene.
Another reveals a koala on a surfboard and the accompanying immediate: “A fluffy koala bear surfs. It has a grey and white coat and a spherical nostril. The surfboard is yellow. The koala bear is holding onto the surfboard with its paws. The koala bear’s facial features is targeted. The solar is shining.”
With audio technology, customers can “create and lengthen sound results, background music or total soundtracks” as much as 45 seconds lengthy, the press launch mentioned. An instance clip of a snake slithering by means of a wooded space consists of the immediate: “Rustling leaves and snapping twigs, with an orchestral music monitor.”
Meta is a bit late to the audio- and video-generation sport as high opponents like OpenAI and Google have already claimed a foothold within the house. OpenAi launched Sora, its video generator, in February, and Google adopted swimsuit with Veo in May.
Meta, nonetheless, has given OpenAI a run for its cash within the AI arms race. Though OpenAI’s ChatGPT debuted first and launched the corporate to worldwide fame, latest iterations of Meta’s Llama mannequin have been nicely acquired. Many seen Llama 3.1, which got here out in July, as superior to OpenAI’s GPT-4o, which got here out shortly earlier than.
Meta says its new “state-of-the-art fashions” outperform opponents in A/B human comparisons. For video technology, these Meta surveyed most popular Movie Gen over OpenAI Sora, the corporate’s press launch mentioned. Meta did not share an A/B comparability with Google’s Veo, which additionally provides sound results and music, however Meta mentioned in a prolonged accompanying analysis paper that it believes Google’s video-to-audio technology fashions could also be extra restricted in size than Meta’s.
Meta, OpenAI, and Google didn’t instantly reply to a request for remark.
