OpenAI releases Sora, artificial intelligence that creates incredible videos with text commands

A. OpenAI, responsible for platforms such as ChatGPT and DALL-E, presented this Thursday (15) a new generative artificial intelligence. This soraIt is a service that has the ability to create videos from text commands.

The language model allows the creation of videos in many different styles. You can ask the system to create photorealistic clips with a retro aesthetic, in animation format, or to simulate a certain genre of cinema, for example.

The same strategy for commands applies to other AIs: The more specific and detailed the prompt, the better. In Sora’s case, it is possible to identify characters, scenery, setting, aesthetics, objects in the scene, and even a sequence of actions.

Prompt: “A movie trailer featuring the adventures of a 30-year-old spaceman wearing a red fleece knit motorcycle helmet, blue skies, salty desert, cinematic style, vibrant colors shot on 35mm film.” pic.twitter.com/0JzpwPUGPB

— OpenAI (@OpenAI) February 15, 2024

OpenAI released a series of clips made using Sora; including the command prompt used to create each of these. On the company’s website, it is possible to see several examples of videos in a wide variety of styles, all artificially created by the new platform.

How does Sora AI work?

The biggest difference between Sora and other language models is its ability to understand. “How some things come to exist in the physical world”. In other words, AI knows the location, context and behavior of elements in reality and can replicate all of this in clips.

The model also includes: “A deep understanding of language” interpreting commands and even creating humans that express emotions. For example, in the video below there was this command:

Prompt: “A stylish woman walks down a Tokyo street filled with warmly glowing neon and bustling city signs. she is wearing a black leather jacket, a long red dress and black boots, and she is carrying a black bag. She is wearing sunglasses and red lipstick. she walks confidently and comfortably… pic.twitter.com/cjIdgYFaWq

— OpenAI (@OpenAI) February 15, 2024

“A stylish woman walks down a Tokyo street filled with warm, bright neon and animated city signs. She wears a black leather jacket, a long red dress, black boots, and carries a black bag. She wears sunglasses and red lipstick. She walks.” confidently and comfortably. The street is damp and reflective, the colored lights create a mirror effect. “A lot of pedestrians are walking there.”

According to OpenAI, Sora can also be used to “extend” clipsis extending their runtime with new AI-powered content based on the original material.

https://t.co/rmk9zI0oqO pic.twitter.com/WanFKOzdIw

— Sam Altman (@sama) February 15, 2024

For now, it’s only possible to create videos of a maximum length of 60 seconds using the service, without any type of audio being automatically connected. The clips also have some flaws, such as deformation of limbs and failure to interact with certain objects.

When can I use Sora?

OpenAI for now It has not confirmed how and whether AI will be rolled out to all users. The clips released so far come from the platform’s experimental phase, which will now undergo a security and privacy assessment looking for “critical areas due to damage or risk.”

Besides, the company has made Sora available to “a range of visual artists, designers and filmmakers” to get input on how to improve the model before commercial release.