Sora
By OpenAI
Sora is OpenAI's text-to-video generation model that creates realistic and imaginative video clips from written descriptions, and can also extend, edit, or blend existing video and image inputs.
Definition
Sora is OpenAI's text-to-video generation model that creates realistic and imaginative video clips from written descriptions, and can also extend, edit, or blend existing video and image inputs.
Overview
OpenAI first showed Sora publicly in February 2024, demonstrating notably longer and more coherent video clips than most prior text-to-video systems, along with the ability to simulate physical detail and camera motion described in a prompt. It later became more broadly available as a standalone product, extending OpenAI's generative lineup beyond text (ChatGPT) and images (DALL-E) into video. Like image diffusion models, Sora generates video by refining noisy visual patches over multiple steps guided by a text prompt, but extends this into the time dimension so that motion stays visually consistent across frames. It supports generating clips from text alone, animating a still image, or extending and remixing existing video. Sora competes with other generative video tools such as Runway and Kling AI, and its release intensified public debate over deepfakes, misinformation, and copyright — themes explored further in Multimodal AI: Vision, Audio, and Beyond and AI Safety and Ethics.
Key Features
- Text-to-video generation from written prompts
- Image-to-video animation from a single still image
- Video extension, remixing, and editing of existing clips
- Relatively strong temporal consistency of motion and objects across frames
- Simulated camera movement and scene composition guided by prompts
- Content moderation and provenance features such as watermarking
- Integration into OpenAI's broader generative product lineup