OpenAI has officially launched Sora.
On Monday, CEO Sam Altman immediately kicked off the livestream by announcing the Sora public release. As of today, Sora will be available today to ChatGPT Plus and Pro users in the U.S. and other countries — excluding the UK and countries within the EU.
Sora reportedly shipping as part of ’12 Days of OpenAI’ livestream marathon
Early last year, OpenAI introduced the AI video generator. But it has been quietly in development to a select group of testers, until being unleashed to the public today, allowing plenty of hype to build up in between. Early demos of Sora’s photorealistic details and advanced sense of physics and accuracy has wowed AI enthusiasts. But it has also drawn controversy over how it got so good. OpenAI hasn’t revealed what Sora was trained on, but reportedly transcribed over a million hours of YouTube videos, suggesting the AI video was trained on videos from the web without the credit or compensation of the video creators. Recently, a group of artists working as early testers for Sora leaked the API credentials in protest of what they call “art washing.”
Perhaps, in an indirect response to criticisms that tools like Sora are exploiting and replacing the work of creatives, Flynn emphasized that “Sora is a tool” and an “extension for the creator behind it.”
Mashable Light Speed
In the livestream, OpenAI product lead Rohan Sahai and product designer Joey Flynn wasted no time in sharing Sora’s capabilities. The tool lives on a standalone website, sora.com, with an explore tab for discovering what other users are creating. By clicking into one of the videos, users can see the methods used to the create the video, such as simple text prompt, video or image extension, or Sora’s storyboarding tool.
In the library tab, users can get started making their own video with a text prompt or uploading an image. From here, users can select aspect ratio, resolution up to 1080p, duration up to 20 seconds, and multiple variations of the prompt. There are also certain default presets like “stop motion” and “balloon world.”
Sora also comes with a more advanced tool called Storyboard, which allows users to shape the video with specific directions. Storyboard bears a similar resemblance to other video editing tools with frame views on the bottom and various editing tools. Each “storyboard card” or frame can be generated from a text prompt or image upload. Users can use the recut feature to shift cards around, the remix feature to describe specific changes to the sequence, the loop tool to play the video on repeat, or blend to create a transition between multiple scenes.
In OpenAI’s announcement, the company shared some of its safety measures. All Sora-generated videos come with C2PA invisible watermarks, and visible watermarks by default. Sora will block “damaging forms of abuse, such as child sexual abuse materials and sexual deepfakes,” and limits uploads of people.
ChatGPT Plus users get 50 videos a month at 480p (or 720p for less videos) and ChatGPT Pro users get 10 times more usage.
Topics
Artificial Intelligence
OpenAI