Meet 'Sora', OpenAI's text-to-video AI model; know how it works

The tech industry is rapidly changing, with advancements like text generation in tools such as chatbots and the emergence of video generation from text prompts.

OpenAI, a leading AI company, has unveiled Sora, a new software capable of creating hyper-realistic one-minute videos based on text prompts.

Currently, Sora is in the information gathering phase, where OpenAI is identifying flaws in the system.

The company is collaborating with visual artists, designers, and filmmakers to gather feedback.

Sam Altman, CEO of OpenAI, introduced Sora on his X account, sharing various videos to demonstrate its capabilities.

here is a better one: https://t.co/WJQCMEH9QG pic.twitter.com/oymtmHVmZN

— Sam Altman (@sama) February 15, 2024

While the software is still undergoing testing, there's no news about its wider release.

What exactly is Sora?

According to OpenAI, it's a text-to-video model that generates one-minute videos while maintaining visual quality and adhering to the user's prompt.

Also Read | How to Use AI Tools Without an Internet Connection on Your PC, Know Here

Sora can create complex scenes with multiple characters and specific motions, capturing accurate details of subjects and backgrounds.

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

— Sam Altman (@sama) February 15, 2024

The model understands not only the user's prompt but also how it translates into the real world.

Altman showcased Sora's creations based on follower requests, featuring diverse scenarios like cycling dolphins and a squirrel riding a dragon. 

Essentially, Sora is a diffusion model capable of generating entire videos at once or extending them to make them longer.

https://t.co/qbj02M4ng8 pic.twitter.com/EvngqF2ZIX

— Sam Altman (@sama) February 15, 2024

It utilises a transformer architecture similar to GPT models, offering superior scaling performance.

OpenAI emphasises that Sora builds upon research from projects like DALL-E and GPT Models, boasting an in-depth understanding of language to interpret prompts accurately and create characters with vibrant emotions.

welcome to bling zoo! this is a single video generated by sora, shot changes and all. https://t.co/81ZhYX4gru pic.twitter.com/rnxWXY71Gr

— Bill Peebles (@billpeeb) February 15, 2024

Additionally, OpenAI plans to include C2PA metadata in the future for further safety.

While other tech giants like Google and Meta have also explored text-to-video models, OpenAI's Sora seems to have made significant strides in this field.