OpenAI Launches Early Access on Text-to-Video AI Model for Red Teamers, Visual Artists

OpenAI introduced its latest AI model that can convert textual prompts to one-minute-long videos, allowing visual artists to generate complex scenes with multiple characters and specific types of motion.

For feedback, Sora, OpenAI's new AI model, will be available to red teamers, designers, filmmakers, and visual artists.

OpenAI Text-to-Video Sora

OpenAI revealed that Sora can generate complex scenes with multiple characters, accurate details of the subject and background, and specific types of motion.

Visual artists can create scenes using Sora by simply typing a prompt. For instance, users can instruct the AI model to create a scene where a camera follows behind a white vintage SUV with a black roof rack.

According to OpenAI, the prompt can be as specific as it can get and Sora will generate every instruction to create the scene.

"The model has a deep understanding of language, enabling it to accurately interpret prompts and generate compelling characters that express vibrant emotions," the company announced.

OpenAI Sora Still in Early Stage

Sora is currently in the early stage of its progress. OpenAI will only open it to select visual artists who will provide feedback on how to improve the model and how it could be helpful to the creative industry.

In addition, red teamers will also test the AI model on the critical areas of it. The team will be assigned to observe potential harms or risks that the model might encounter.

"We're teaching AI to understand and simulate the physical world in motion, with the goal of training models that help people solve problems that require real-world interaction," the company added.

However, the current model is still struggling to accurately project the physics of complex scenes. The problem could be obvious in scenes that need continuity. For instance, biting on food might not be reflected in the next scene.

There could also be some mix-up on left and right, along with following some precise descriptions of events like following a camera trajectory.

Regardless, OpenAI assured the public that the company is planning to include C2PA metadata in the future so that any video generated by Sora will be tagged properly as AI-generated content.