OpenAI Unveils Sora Video Generator
OpenAI has launched Sora, a generative AI model that creates high-quality, minute-long videos from text prompts. In a rapid adoption, Microsoft is already offering free video generation with Sora via its Bing platform. Additionally, OpenAI has reached an agreement with Disney to permit the use of its characters on the platform, setting a precedent for licensed, AI-assisted content.
- OpenAI's Sora was first unveiled on February 15, 2024, and can generate videos up to a minute long from text prompts, create videos from still images, and extend existing video clips. - The model uses a transformer architecture, similar to GPT language models, and a diffusion model that starts with a frame of static noise and gradually refines it to create the final video. - Public access to Sora began on December 9, 2024, for ChatGPT Plus and Pro subscribers in the U.S. and Canada, with limitations on video resolution and the number of videos that can be generated per month. - To address safety concerns, OpenAI has implemented measures such as visible watermarks on all generated videos, C2PA metadata to indicate AI origin, and restrictions on creating depictions of real people to prevent deepfakes. - There has been significant controversy regarding Sora's training data, with OpenAI executives being elusive about whether copyrighted materials from platforms like YouTube were used without permission. - In addition to text-to-video, Sora's features include the ability to animate static images, create looping videos, and blend two different videos into a seamless clip. - Key competitors in the text-to-video space include Google's Veo, Meta's Make-A-Video, Runway's Gen-2, and Pika Labs. - The initial public release does not include availability in the European Union or the United Kingdom.
