OpenAI unveiled its text-to-video generative AI tool Sora last month, capable of producing a clip of up to 60 seconds from text prompts or single images. Though at the time, the company announced that it was only made available for red teamers, as well as specific visual artists, designers and filmmakers for feedback. Now, a company executive has said that the general public can get access to it within the year.
The Wall Street Journal reports that “OpenAIwon’t publicly release Sora, its new text-to-video tool, until later this year”, as soon as a few months, even. In an interview with company chief technology officer Mira Murati, she says that in the meantime, the tool will continue to go through the red teaming process, as well as optimisations to make it “available at similar costs” to DALL-E.
Of course, while the initial announcements come with pretty impressive examples of what Sora can do, when it actually gets to people’s hands, it would be a lot easier to break than the impression that OpenAI gives. The report itself consists of a few examples, impressive as they are at a glance.
But when Sora gets good enough to evade even the most detailed scrutiny – which will be awhile – these generated videos will come with the sort of watermark that DALL-E recently got. But as mentioned at the time, these can be easily removed, probably even accidentally at times.
(Source: WSJ)
