OpenAI Introduces Sora, the AI Model Generating Realistic Videos Based on Input Instructions

Berfin Ceren Meray
Onedio Member
0

Step into the era of video with artificial intelligence! OpenAI, the developer behind ChatGPT, unveils its latest bot, Sora. With this new AI model, the capabilities extend beyond static images, allowing the generation of dynamic scenes and videos of specific lengths based on input instructions. Explore the exciting possibilities as Sora takes AI creativity to a new level in the realm of moving visuals!

OpenAI Introduces "Sora," a New AI Model Capable of Simulating the Physical World in Videos Based on User Instructions

OpenAI, the creator of the ChatGPT model, has unveiled a new artificial intelligence tool named Sora. This cutting-edge bot can simulate the physical world in videos for up to a minute, responding to user prompts regarding both the subject and style.

Sora, inspired by the Japanese word for "sky," showcases its ability to generate realistic visuals up to a minute long, adhering to user instructions on both the subject and style.

The company provides detailed demonstrations on its website, highlighting Sora's capacity to create videos based on still images or expand existing footage with new elements. 🎬 This technology can even resurrect extinct animals from thousands of years ago.

The shared blog post expresses OpenAI's goal: "We aim to build models that help people solve real-world problems requiring real-world interaction. ''

''We're teaching artificial intelligence to understand and simulate the moving physical world.'

An example video created based on the input: "A movie trailer featuring a 30-year-old astronaut wearing a red knit motorcycle helmet, cinematic style, shot on 35mm film, vivid colors, set against a blue sky and a salt desert."

The resulting video was truly astonishing, featuring not only the requested imagery but also the cinematography created by artificial intelligence.

OpenAI has granted access to Sora for several researchers and video creators.

According to the blog post, experts will test the product, referred to as the 'red team,' to assess whether it violates the service terms prohibiting 'extreme violence, sexual content, hateful images, likeness of celebrities, or others' intellectual property rights.'

While access to Sora is limited to researchers, visual artists, and filmmakers, CEO Sam Altman responded to user questions on Twitter with video clips created by Sora.

These videos carry a watermark indicating they were generated by artificial intelligence.

OpenAI previously released DALL-E, an image generator, in 2021, and ChatGPT, a highly productive AI chatbot that rapidly reached 100 million users by November 2022.

While other AI companies have introduced video creation tools, these models could only produce a few seconds of footage, often with minimal correlation to input commands.

Google and Meta have mentioned their ongoing development of productive video tools but have not disclosed details to the public.

For ChatGPT, these concerns are now addressed. The company announced a memory expansion experiment for ChatGPT, allowing users to more comprehensively record their conversations with the AI.

OpenAI has not disclosed the amount of visual data used to train Sora or the source of training videos.

In a statement to The New York Times, the company mentioned that the database includes both publicly available sources and licensed videos from copyright holders. OpenAI has faced copyright infringement claims several times for its AI tools imitating data from massive datasets scraped from the internet. Navigating this complex issue remains challenging, but it is hoped that humanity will continue to benefit from advancements in this new era, with artificial intelligence playing a significant role.