IndexMapAbout
Model FamilyVideoCross-Modal

Sora

Also known as: OpenAI Sora · Sora Turbo

Plain English

OpenAI's text-to-video generation model that creates coherent, multi-shot video clips from prompts by treating video frames as spacetime patches inside a Transformer.

Technical Overview

Sora applies a Diffusion Transformer (DiT) directly to compressed 3D spacetime latent patches produced by a visual VAE encoder. Treating variable-resolution, variable-duration video as unified token sequences enabled long-range temporal consistency, camera motion, and multi-shot framing—catalyzing the industry shift toward DiT-based Video Generation Models (Sora, Veo, Kling, Wan).

Chronological Entries (2)

#2Release
AI-Assisted

OpenAI launched Sora Turbo publicly at sora.com for ChatGPT Plus and Pro subscribers, adding storyboard timelines, remixing, and variable aspect ratios up to 1080p resolution.

Direct Relations (5)