Core Concepts
OpenAI introduces Sora, a text-to-video model, aiming to simulate the physical world in motion and solve real-world interaction problems.
Abstract
OpenAI unveiled Sora, a tool that generates videos from text prompts. Named after the Japanese word for "sky," Sora can create realistic footage based on user instructions, extend existing videos, and adhere to specific styles. The company allows limited access to researchers and creators to test the product's compliance with terms of service. Sora aims to revolutionize video generation by producing detailed scenes with vibrant emotions.
Stats
The new model can produce realistic footage up to a minute long.
OpenAI has not disclosed how much footage was used to train Sora.
Quotes
"We’re teaching AI to understand and simulate the physical world in motion." - OpenAI blogpost