Luma AI launched its artificial intelligence (AI) text-to-video era mannequin, Dream Machine, globally on Wednesday. The platform can generate as much as five-second-long movies from easy or descriptive textual content prompts. The AI mannequin can generate movies in varied types reminiscent of cinematic, animation, practical, and extra. The AI agency claims that Dream Machine was skilled totally on movies, and is able to producing “bodily correct, constant and eventful photographs.” The platform is at the moment free to entry and use, however it’s more likely to have a every day era restrict.

Luma AI debuts Dream Machine

In response to the website, the Dream Machine AI mannequin is constructed on a transformer mannequin and was skilled instantly on movies. Normally, massive language fashions (LLMs) are skilled on textual content and pictures after which are moved to movies, as they require a deeper spatial and movement understanding. “Dream Machine is our first step in the direction of constructing a common creativeness engine,” the corporate added.

Dream Machine joins video era platforms reminiscent of Runway AI and Pika 1.0 that are additionally accessible within the public area and supply video era between three to 5 seconds. Devices 360 tried the platform and located that the immediate adherence of the platform is sub-par. It struggles with a number of characters or when the immediate is simply too advanced. Nevertheless, in comparison with the opposite two, it’s able to producing higher-quality cinematic movies.

The AI platform takes 120 seconds to generate a video which the corporate claims could have 120 totally different frames. Dream Machine can be mentioned to know how folks, animals, and objects work together with the bodily world and create movies with correct physics and character consistency.

Nevertheless, Luma AI additionally highlighted a number of limitations within the present mode reminiscent of motion, textual content, morphing, and the well-known Janus drawback, the place as an alternative of exhibiting a constant 3D output, the AI mannequin reveals a number of canonical views of an object in several instructions.

Luma AI’s Dream Machine creates a video about Stormtroopers
Photograph Credit score: Luma AI/Dream Machine

Luma AI has not shared any technical particulars concerning the AI mannequin, so details about parameter dimension, benchmarks, structure, and coaching strategies is just not identified. The corporate additionally didn’t share any particulars concerning the procurement of the coaching information. Notably, Devices 360 was in a position to generate a number of movies with copyrighted characters in them.

To check out the platform, lovers can go to the web site and click on on the ‘Strive Now’ button. Customers could have to enroll in the platform earlier than they’ll generate movies.