The final decision on this matter has not yet been made.
The Sora neural network from OpenAI may be able to generate explicit videos, admits the startup’s technical director, Mira Murati. According to Murati, OpenAI has not yet decided how Sora's capabilities will be limited.
Sora is a video generator based on test descriptions, shown by OpenAI in mid-February 2024. The model generates videos of fairly high quality: thanks to an increase in the number of parameters, OpenAI was able to get rid of a lot of characteristic glitches such as inconsistency of scenes.
And here's another story
Now, Murati says, artists and content creators are experimenting with Sora. With their help, OpenAI intends to understand where the boundaries need to be drawn in the capabilities of the neural network.
Mira Murati
CTO of OpenAI
I'm not sure [будет ли Sora генерировать нюдсы]. One can imagine a creative environment in which artists would want to have more control in this area. Right now we are collaborating with artists and creators from different fields, trying to understand how flexible the tool should be and what might be useful.
OpenAI cannot yet say when exactly Sora will become available not only to specially invited researchers. The company only promised to release the model before the end of 2024, without telling what exactly the release would look like.
Judging by the first studies, in some cases, videos generated by Sora can mislead some viewers: they mistake the generated videos for real videos. Any explicit videos created with the help of Sora have not yet been shown.
This is interesting