The Wall Street Journal published an interview on 13 (today) local time with the OpenAI Interview with Mira Mulati, Chief Technology Officer. She stated.Sora will be launched this year, "probably in a couple months."
In addition to making Sora openly available to the public, OpenAI plans to add audio generation capabilities that may make the scenes more realistic, Mulati said. Also.The company also plans to allow users to edit Sora-generated content themselves, "because AI tools don't always generate accurate results". She said the company is trying to find ways to enable users to edit and create audio and video content on their own.
When asked what data OpenAI used to train Sora, Murati didn't give a very specific answer: "I'm not going to go into the details of the data used," he said.However, these data are publicly available or licensed data." At the same time, she wasn't sure if videos from YouTube, Facebook, and Instagram were used, saying only that the content Sora used came from Shutterstock because of their partnership.
In addition, Murati said that Sora, after the release of"Likely" to not be able to generate audio/video containing public figuresThe videos will also come with AI watermarks.
Three members of the Sora core team were recently on the WVFRM podcast, and they also said that Sora is still in the feedback phase, "It is not a product yet and will not be available to the public anytime soon.. "
Not long ago, on March 10th, Pika, another text-to-speech video platform, announced a new feature that allows users to seamlessly generate and integrate sound into Pika-generated videos.You can describe the desired sound with a cue word, or just let Pika automatically generate the sound based on the video content..