OpenAI says Sora 2 can generated Western-style footage using a simple prompt like "a person is standing on 2 horses with legs spread. make it not slowmo also realistic. the guy fell off pretty hard in the end. single shot." — Photo: OpenAI/dpa
SAN JOSE: AI software has long been able to generate videos from text prompts, but ChatGPT developer OpenAI now wants to make it easier to create fun, shareable clips by letting users put their own faces in videos.
Once users film themselves with audio in the new OpenAI app, Sora, the software can then depict them in any scenario they specify in a prompt.
OpenAI gives examples of videos created from prompts like "a man does a backflip on a paddleboard" and "figure skater performs a triple axle with a cat on her head."
The company's demo videos are often of a bizarre or cinematic nature with unrealistic movements and scenarios, but at times also appear to resemble genuine footage.
The Sora app, initially launching only in the US and Canada, is also designed to function as an online network for AI-generated videos. Users can not only view and edit clips created by others but also include friends registered on Sora in their own videos.
OpenAI demonstrated the app's capabilities with a video featuring a digitally generated version of chief executive Sam Altman walking through a virtual world and standing on the Moon.
The first version of the Sora model was released last year. According to OpenAI, Sora 2 is now better at generating scenarios with complex movements, such as gymnastics.
According to The Wall Street Journal, OpenAI plans to take a new approach to the use of copyrighted content and characters in the Sora software. The company will ask rights holders, such as film studios, to explicitly object to the use of their material — otherwise, it may be included in videos.
Rights holders have repeatedly accused AI companies of using their intellectual property without permission to train their software.
At the same time, OpenAI has stated that the Sora software will not replicate the likeness of public figures without their consent, The Wall Street Journal reported, citing informed sources. – dpa
