Sora, the OpenAI’s text-to-video generator will be made public later this year

After ChatGPT, the tech world is now ready to embrace another major AI model. Just recently, OpenAI revealed Sora, its text-to-video generative AI model. Right now, the new AI tool is only accessible to members of the red team. According to some recent pieces of information, the company might soon launch the tool to the general public.

Mira Murati, the chief technology officer (CTO) at OpenAI, shared the future plans regarding Sora in an interview with The Wall Street Journal. Murati stated in the interview that the AI tool might be completed in a few months and that it will be released “this year.” The CTO claimed that the company wants the videos generated by Sora to be editable. With time, the company will add audio into AI-generated videos.

Murati mentioned in the interview that Sora is quite expensive than other AI models. However, according to reports, OpenAI wants to make the technology “available at similar costs” to its text-to-image AI tool, DALL-E. Similar to other gen AI tools, Sora require data for training. Murati didn’t share any accurate details regarding this. However, it was mentioned that publicly available and licensed data has been used for training Sora. She acknowledged that Sora pulls content from Shutterstock, with which OpenAI has a collaboration, but she couldn’t say if Sora also pulls from Facebook, Instagram, YouTube, or other sources.

As a protective measure, the videos generated by Sora will carry a watermark. The new AI tool is perhaps incapable of producing images of public figures. It’s unclear if this will be sufficient to stop misinformation, since watermarks have been cracked in the past.