What if you could just write what you wanted to see, and then an AI generated a realistic video for you in real time?
Now you can get exactly that from a new feature OpenAI just revealed.
In an announcement on Thursday, February 8, 2024, OpenAI announced “it has expanded beyond text and images to offer video-generation AI for the first time.” according to an article on CNBC.
OpenAI’s new AI model is called Sora, and it allows a user to type what they’d like to see in a video scene, and the software then generates a photo-realistic high-definition video rendering in real-time.
OpenAI already has an image-generation AI tool called DALL-E, but the introduction of Sora takes video generation to a whole new level.
“Sora can also generate video clips inspired by still images, and extend existing videos or fill in missing frames.” according to the same article on CNBC.
With this announcement instantly came fears of even higher, better quality deep fake videos, especially in the lead up to the 2024 Presidential race here in the U.S., although any AI generated video of either of the front-running candidates running would be easy to spot as a fake due to the candidates ages.
OpenAI is already on top of this.
They have a group of safety testers called “red teamers” testing the model for vulnerabilities including for misinformation and bias, and they are building a “detection classifier” that can identify Sora-generated video clips. With this functionality they would be able to identify deep fakes generated with Sora AI-generated videos based on the video metadata from when it was generated.
At least it won’t be a boring future we have to look forward to.
OpenAI’s new software lets you create realistic video by just typing a sentence