Sora
More info
- Creator
- OpenAI
- Released
- Type
- Text-to-video model
- Website
- openai.com
OpenAI's video-generating model — shocked the world in February 2024 with 60-second 1080p clips from text prompts.
Diffusion transformer architecture ("DiT"). Breaks video into "spacetime patches" — analogous to tokens in text LLMs. Public release in December 2024 (with strict filters). Competitors exploded within the year: Runway Gen-3, Luma Dream Machine, Kling (Chinese), Pika, Google Veo, Hunyuan Video (Tencent, open weights). Uses: storyboards, social-media content, indie film, contentious deepfake debate.