OpenAI is Launching Text-to-Video AI Model Sora

OpenAI just announced its newest AI model, a text-to-video model called Sora. And the promo videos are really impressive.

A year ago, the promise of quality AI-generated images and videos seemed far away. Remember the creepy hands epidemic?

But it looks like OpenAI has once again raised the bar and come out with a model that may revolutionize video.

What We Know About Sora

Sora is an AI model created by OpenAI that can generate high-quality video up to a minute long based on a text prompt.

“Sora is able to generate complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background,” OpenAI wrote in a blog post. “The model understands not only what the user has asked for in the prompt, but also how those things exist in the physical world.”

But Sora’s not without issues.

OpenAI says the model can struggle to correctly display cause and effect. For example, say it shows someone drinking water out of a glass, it may not show the water level lowering with every sip.

OpenAI’s engineers are also still working on improving its spatial understanding, things like left and right or forward motion.

Here’s an example provided by OpenAI where Sora shows someone running on a treadmill but instead of running facing the machine, the jogger runs in the opposite direction.

Image Source

The model is not currently available to the public and the company has not shared a release date. It’s currently in a red teaming phase, where specialized users will test it for potential harm.

When you put AI and video together, the danger is massive. We’ve seen recently how quickly deep fake images can spread and how disturbing they can be. Video raises the stakes exponentially.

Here are some safeguards OpenAI has or plans to have in place:

• A detection tool that can alert users when a video was generated by Sora
• C2PA metadata to track the origin of an video

Sora is also designed to reject prompts that violate OpenAI’s content policy – for instance content that illustrates violence, sexual behavior or abuse, and public figures.

In addition to red teamers, a team of visual artists, designers, and filmmakers will also have access to Sora, to provide feedback on how it can enable creative work.

The model’s been met with both excitement and fear, as with everything AI-related. Some are excited about the possibilities for businesses and individuals alike to create quality images with simple prompts.

For businesses with limited creative resources, this could be a game changer.

However, some worry about the impact on the job market for creative roles, like designers and animators. Would their roles now become redundant?

Though that’s unlikely, this fear always rises to the surface whenever a new tool comes around. It’s just too early to know, especially since the model isn’t available to the public yet.

Is it as good as it seems from OpenAI’s promos? We’ll have to wait to find out.