OpenAI has announced the launch of Sora,  the updated version of its audio and video generator Sora that was unveiled last year.

Along with it, the AI startup has also launched a new social iOS app just called “Sora,” powered by Sora 2. The app will initially be rolled out in the US and Canada, and will be expanded to other countries.

In a blogpost, OpenAI said that Sora 2 can do things that are "exceptionally difficult" and in some instances "outright impossible" compared to previous models.

"Olympic gymnastics routines, backflips on a paddleboard that accurately model the dynamics of buoyancy and rigidity, and triple axels while a cat holds on for dear life," are some of the things the new model can achieve, it explained.

Underlining the advancement of the current model, the blogpost further said that prior video models are overoptimistic in that they will morph objects to make it less realistic. However, with Sora 2, for instance, if a basketball player misses a shot, it will rebound off the backboard as opposed to maybe the ball spontaneously teleporting to the hoop for previous models.

The model has also improved in "controllability" as it is able to follow intricate instructions spanning multiple shots while accurately persisting world state, it noted.

It also highlighted its realistic, cinematic, and anime styles excellence.

Notably, users can also directly inject elements of the real world into Sora 2, it said.