OpenAI's generative video tool, Sora, made headlines when it was unveiled last month for generating studio-quality video from text

Speaking to the Wall Street Journal, OpenAI CTO Mira Murati said that Sora will be available to the public by the end of this year, possibly within a few months So far we have only seen pre-built videos of OpenAI

Before it is released, OpenAI is trying to make Sora more efficient and less expensive The goal is to produce video as fast as DALL-E produces images and for about the same price

This suggests that Sora could be bundled with ChatGPT Plus as an API for third-party developers to integrate into their products

Mulati said they are also "red teaming" to find ways to ensure that the AI video generator does not generate objectionable content and minimize the risk of false alarms

News of a possible future release of Sola coincides with a new survey commissioned by Variety that found that most US adults could not tell whether a video was Sola or real

The survey was conducted by HarrisX and included demonstration videos released by OpenAI of waves crashing on steep cliffs and the California gold rush

Many respondents even suggested that human-generated videos were created by AI, suggesting that OpenAI and other video platforms need to do more work on identification

Mulati told the Wall Street Journal that the topic keeps him up at night and that he may consider postponing the release if security measures are not put in place before the November presidential election

It is important to "address the issue of misinformation and harmful bias," Murati explained He added, "We will not release anything that we are not confident about with respect to how it will affect elections or other issues around the world"

Mulati said that when "Sola" is finally released, it will cost about the same price as "Darley" and take a similar amount of time This is because it can currently take "several minutes" to produce a simple 20-second video

Mulati said they are also working on improving functionality, as there are still some impractical issues with Sola, such as finger movement

"ChatGPT and DALL-E are optimized for public use, but Sora is more expensive because it is a research product," Murati told the WSJ

"We don't know what it will look like when it is released to the public, but we hope it will be similar in cost to DALL-E"

"We are trying to figure out how to use this technology as a tool for people to edit and create," Murati explained 'Ultimately, it will be more steerable, more controllable, and able to accurately reflect the intent of what we want' [Open AI is currently in the process of "red teaming" Sola, testing its limits and trying to get it to do less than ideal things, such as generate videos that are illegal, unrealistic, or potentially defamatory in their depiction of public figures

The result is a model with guardrails similar to OpenAIs image model DALL-E For example, it cannot generate videos of public figures

According to Mulati, they are also working with a wide range of artists and policymakers to determine the level of flexibility the tool should provide for things like nudity and violence

"We see it as a tool for extending creativity We want people in the film industry and creators of all disciplines to play a part in informing how the tool can be developed and deployed" 'AI tools will extend our creativity, our knowledge, our collective imagination, and our ability to do just about anything However, finding a path to introduce AI tools into our everyday reality will be extremely challenging along the way, but I think it is worth the challenge"