OpenAI Announces Sora, Its Text-to-Video Model

Sora - OpenAI

OpenAI on Thursday announced its text-to-video AI model called Sora. Sora can generate “complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background,” the company says. At this time, Sora is only available to red teamers (experts in areas like misinformation and bias), and “a number” of visual artists, designers and filmmakers, before being made widely available.

The company showcased some videos generated by Sora and the results are impressive. OpenAI says the text-to-video AI model understands “not only what the user has asked for in the prompt, but also how those things exist in the physical world.” Sora is capable of creating up to one-minute long videos “with multiple shots within a single generated video that accurately persists characters and visual styles.” While Sora can create realistic videos from text prompts, it may struggle with accurately depicting the physics of complex scenes.

In addition to generating videos solely from text-based instructions, Sora can generate videos from a still image while using the image’s theme for accuracy. The AI model can also fill in missing frames from an existing video or extend it.

OpenAI says it is building tools to help detect when a video is generated by Sora and misleading content. The company will also be employing existing safety measures from DALL.E 3 – its image model, in existing products.

There is no date yet for when Sora will be widely available, but it shouldn’t take long. Earlier in the week, the company announced memory, a feature that lets ChatGPT remember conversations.

Image – OpenAI

Posted by Abiodun

Passionate about Technology and everything concerning it. Avid Gamer and Music Lover. Loves Chelsea FC. Overall, a nice guy.

Leave a Reply

Your email address will not be published. Required fields are marked *