What is OpenAI’s Sora and why should you care that it’s officially here? AI video generation by OpenAI has been one of the company’s more anticipated releases in the last year, especially since the company first unveiled the tool to testers in February. The tool’s photorealistic video and smooth graphics are certain to be a turning point in the use of AI, however, there are multiple concerns about its functionality and application.

OpenAI’s Sora has now been released for ChatGPT Plus and Pro users in select regions, which means that we could soon see more sophisticated and realistic AI videos make the rounds in the coming weeks.

OpenAI Sora release

Image: OpenAI’s Sora

What is the New OpenAI Sora Tool? Understanding the New AI Video Generator

OpenAI’s Sora is an AI video generation tool that can be used to create 5 to 20-second videos of up to 1080p resolution. The user can pick the aspect ratio of the generated videos, whether widescreen, vertical, or square. Content can be generated entirely using text prompts but you can also add in your own assets to build on or edit to your preferences.

The foundational text-to-video Sora AI model was initially only released to red teamers who could assess the tool for prime areas of risks, as well as to a limited number of artists and other creatives who could validate the quality of the content generated. 

Right from the start, it was evident that the OpenAI Sora release would be groundbreaking when it finally came to fruition, considering the detailed nature of the content and its ability to respect the laws of nature or those of physics in the content it created. It wasn’t perfect and frequently made errors—some of the generated content remains more jarring than anything a human mind has ever come up with—but it was noteworthy nonetheless. 

The interface and capabilities have also been upgraded nicely, making it easier to navigate and use. Sora’s new Storyboard feature makes it more convenient for users to review and edit their video frame by frame, or blend multiple videos together to create flawless transitions between scenes. You can replace, remove, and remix away to your heart’s content.

Who Can Use OpenAI Sora’s Abilities?

The latest version of the AI, Sora Turbo, is available as a standalone product on the Sora.com website for ChatGPT Plus and Pro users. With a $20 monthly ChatGPT Plus subscription, users can generate up to 50 videos a month at 480p or fewer videos at 720p. The $200 monthly subscription for the Pro Plan gives users “10x more usage, higher resolutions, and longer durations.”

The tool is available to users everywhere that ChatGPT is available, except for the regions in the UK, Switzerland, and the European Economic Area. The company is working on expanding the AI to those areas in the upcoming months. 

ChatGPT Team, Enterprise, or Edu users will not gain access to the AI, so those who want to test it out may have to consider getting a separate Pro or Plus subscription to do so. 

OpenAI Sora video generator

OpenAI Sora’s Features And Capabilities Are Worrying Some Users

Once you understand what OpenAI’s Sora is, there are often two common reactions to the tool—excitement at the possibilities or fear over misuse of the AI. The Sora video generator is undoubtedly one of the most adept AI video tools we have right now. Other software like Google’s DeepMind Genie 2 has made some impressive progress as well, but these are still restricted to limited use cases and to specific audiences.

OpenAI’s Sora release is still restricted to its paid subscribers, so we won’t see it used everywhere. However, as the number of people with access to these tools continues to increase, so do the possibilities of misuse. Aware of this, OpenAI has attempted to restrict the kind of content that can be created using the tool.

Child sexual abuse material (CSAM) and sexual deepfakes are blocked from the service and any attempts to create such content are to be reported to the National Center for Missing & Exploited Children (NCMEC). The videos generated also contain C2PA metadata, or watermarks, to create a clear way to distinguish them as AI-generated.

Additionally, one key area of potential misuse is the creation of content that imitates real people. Likeness is only available as a pilot feature to limited testers and there are checks in place to moderate content with people in it, with blocks on any nudity. Such safeguards can help dissuade misuse, however, users are not convinced that this is efficient enough to protect everyone from AI.

Concerns About Sora’s Training Material Abound

Sometimes a service is too good to be true and that’s how many feel about Sora’s features and the data used for its training. Some of the content generated using the AI is so realistic that it now has users worried about how exactly the AI is being trained to be this efficient. Unfortunately, OpenAI has kept mum about its training data and just what material it has used to perfect its AI.

Instead of purchasing or licensing data from creators and artists, OpenAI and its competitors have been accused of transcribing hours of videos from YouTube to train its AI tools. All ML models have always capitalized on content available on the web without permission from or compensation to any of the original creators, so this news is unsurprising. However, their lack of transparency continues to be a problem.

While users worry about the misuse of AI tools for malicious purposes, we may have to first overcome the hurdle of AI companies misusing data without consequence. With the latter issue unlikely to ever see a solution, the use of AI continues to be an uncomfortable experience for many. 

tagged in AI video generation by OpenAI, OpenAI Sora features, OpenAI Sora release, OpenAI Sora video generator, what is OpenAI Sora
Share: