What is Sora?
Sora is a state-of-the-art text-to-video system from OpenAI that extends the diffusion-transformer approach used in DALL-E 3 to the temporal domain.
Key capabilities
- Generates 1080 p video up to 20 s from natural-language prompts.
- Accepts images or existing clips as starting points, enabling remix, extend, loop, blend & storyboard workflows.
- Ships with community style-presets and a responsive timeline editor for fine-grained control.
Release timeline
- 15 Feb 2024 – first public preview with sample videos and technical report.
- Dec 2024 – rolled out to ChatGPT Plus & Pro subscribers.
- 2025 – integrated tightly with ChatGPT’s multimodal interface and expanded to additional regions.
Limitations & safeguards
OpenAI flags Sora videos with C2PA provenance metadata and blocks prompts that request disallowed sexual, violent, hateful or celebrity content. Physical-world reasoning still shows occasional errors in complex scenes.
Who is behind it?
Sora is developed by OpenAI, leveraging research contributions from the same team that produced GPT-4o and DALL-E 3.
