Apex Tech
What is Sora?

OpenAI stuns again with Sora, an AI video generator producing eerily lifelike content. Its diffusion model with transformer architecture hints at AI’s future. Explore what’s known and why public release may be delayed.

What is Sora?

Sora, developed by OpenAI, is an AI model specializing in text-to-video generation. This means it can transform text prompts into short video clips. The latest batch released by OpenAI is remarkable, showcasing Sora’s ability to create intricate scenes with multiple characters, precise motions, and detailed backgrounds. According to OpenAI’s recent blog post, Sora comprehends not only the user’s prompt but also how those elements manifest in the real world. The model’s proficiency in generating lifelike videos marks a significant advancement in AI technology.

Who can use sora?

Sora is currently undergoing rigorous testing by security researchers to ensure its safety and security before its public release. These researchers are assessing “critical risks” associated with the AI model.

Additionally, a select group of visual artists, filmmakers, and designers have been granted access to Sora for trial purposes. However, the specific participants in this trial have not been disclosed.

Information circulating on the OpenAI forum suggests that a waiting list may be introduced at some point, offering the first opportunity for users to access Sora. Unfortunately, there is no indication of when this sign-up process will become available.

Why the delay with Sora?

As previously discussed, OpenAI has postponed the release of Sora to the public, citing concerns about the thoroughness of safety testing for the video-generating technology. This cautious approach is warranted, given the multitude of ethical issues surrounding the creation of realistic videos, especially during periods such as election years.

“We will be implementing several critical safety measures before Sora is integrated into OpenAI’s products,” the company stated. “This includes collaboration with red teamers—experts in areas such as misinformation, hateful content, and bias—who will rigorously test the model.”

Furthermore, OpenAI is developing an AI video detection classifier capable of discerning whether a video has been generated by Sora. This strategy mirrors a similar initiative undertaken following the launch of ChatGPT, where OpenAI introduced a text classifier. However, this classifier was eventually discontinued due to its lack of reliability. In our testing, it failed to detect content generated by ChatGPT itself, including instances of potential plagiarism.

Working Model of Sora

“Sora is a diffusion model that generates videos by initially presenting them as static noise and gradually refining them over multiple steps,” OpenAI explained this week.

This technology operates similarly to the GPT series of language models used in OpenAI’s ChatGPT chatbot. They all utilize a “transformer” architecture—a type of neural network that takes inputs and transforms them into outputs. Additionally, elements of DALLE-3, such as the recaptioning system, have been incorporated into Sora.

OpenAI has provided Sora with videos and images as units of data, known as “patches.” “By standardizing how data is represented,” OpenAI explains, “we can train diffusion transformers on a broader range of visual data than was previously possible, encompassing different durations, resolutions, and aspect ratios.”

For further insights into OpenAI’s AI video generation technology, you can explore a blog post on the company’s research portal.

When will it be available for us?

Sadly, there is currently no release date for Sora. All the content that’s been going viral on the internet over the past 24 hours has come out of the announcement blog post published by OpenAI.

Interestingly, it doesn’t seem like OpenAI has even given a vague indication of when it might be made generally available – there’s not even been an indication that it will be released this year.

That’s quite unusual for such a large announcement, and could suggest it’s a long way off public release – but then again, OpenAI does admit that it’s sharing its research early. With the speed that the artificial intelligence industry has moved over the past two years, the true launch date is anyone’s guess.

Also Read