1. Home
  2. Economic News

Openai's Video Model Sora Has Captured Users' Interest.

OpenAI's latest artificial intelligence video tool, Sora, has impressed social media users with its realism, despite not yet being ready for public release.

On Thursday, the artificial intelligence (AI) company OpenAI excitedly announced its first text-to-video generation model, although the company acknowledged areas where the model still needs development. OpenAI named the new AI model Sora, introduced on February 15th. Sora is said to be capable of creating detailed videos from simple text commands, expanding existing videos, and even generating scenes based on still images.

OpenAI claims that the artificial intelligence model can generate film-like scenes in resolutions up to 1080p. These scenes can include multiple characters, specific types of movements, and accurate details of subjects and backgrounds.

The working principle of Sora is based on a diffusion model, similar to OpenAI's image-based models like Dall-E 3. Diffusion involves creating an image that resembles "static noise" through an image-based model and then gradually removing the "noise" in several steps to obtain a clear image output.

The artificial intelligence company stated that Sora builds upon previous research such as ChatGPT and Dall-E 3, allowing the model to better reflect user inputs.

OpenAI acknowledged that Sora still has some weaknesses and may struggle to accurately simulate the physical structure of a complex scene, potentially mixing up cause-and-effect relationships:

"For example, a person may take a bite from a cookie, but later there may be no bite mark on the cookie." The company also noted that the new tool may mix up the "spatial details" of a given command by also mixing up left and right or failing to follow precise definitions of directions.

OpenAI stated that the new model is currently only available for use by selected designers, visual artists, and filmmakers who are tasked with assessing "critical areas for harm and risk" and providing feedback on how the model can be improved.

A report from Stanford University in December 2023 revealed that AI-powered image generation tools using the AI database Laion were trained on images of illegal child exploitation. This raised serious ethical and legal concerns for models that generate images or videos from text.

Users on X were amazed While dozens of demos showcasing Sora's working examples circulated on X, Sora became a trend on X with over 173,000 posts.

In order to demonstrate what the new model is capable of, OpenAI CEO Sam Altman accepted custom video creation requests from users on X and shared a total of seven videos created by Sora, ranging from a duck on a dragon's back to dogs recording podcasts on a mountain top.

Artificial intelligence expert Mckay Wrigley expressed his surprise at the video created by Sora.

Jim Fan, senior researcher at Nvidia, stated in a post on X on February 15th that anyone who believes Sora is just a "toy" like Dall-E 3 would be greatly mistaken.

According to Fan, Sora is more than just a video creation tool because the artificial intelligence model not only generates abstract videos but also decisively determines the physical properties of objects in the scene.

You can follow the latest developments and news in the cryptocurrency markets in real-time on Kriptospot.com.

You May Also Like