OpenAI Sora
by openai
Text-, image-, and video-driven generative video model (preview)
See OpenAI Sora announcement / docs
Summary
Sora is OpenAI’s advanced text-to-video / image-to-video / video-to-video model, designed to generate up to ~60 seconds of coherent video from natural language prompts and/or input media. Sora uses a diffusion-transformer architecture with space-time patch representations and a video compression/latent pipeline to scale across durations and aspect ratios.
Features
- Text → video generation (complex scenes, multi-character, sustained coherence)
- Image-to-video and video-to-video (can place input media at designated points in output)
- Up to ~60 seconds in previewed demos (subject to change)
- Asynchronous job-based API: create jobs, poll status, download results
- Strong language grounding for cinematic and narrative prompts
- Active red-team / safety testing during preview rollout
Superpowers
Useful for creatives and teams who want to convert scripts or visual briefs into short, realistic video clips quickly. Sora excels at generating narrative sequences with multiple elements, offering a good balance of fidelity and conceptual understanding for film, education, and marketing prototyping.
Pricing
- Preview / limited access; pricing and broad availability not generally public during preview. Check OpenAI’s product pages and API docs for updates.
Known limitations & safety
-
Preview status: access restricted, with OpenAI conducting safety/red-team evaluation.
-
Limitations in spatial reasoning and complex causality persist in current-generation video models.
-
Expect content policy restrictions and monitoring for misuse.
Sora 2
Sora 2 is an incremental update announced in late 2025 that improves temporal coherence, visual fidelity, and native audio synchronization over the initial Sora preview. It extends Sora’s capabilities for longer, higher-fidelity clips and tighter lip-syncing for characters, and it added improved guardrails during the preview rollout. Update details and API changes are tracked in OpenAI’s developer changelog and blog posts — see OpenAI docs for the latest rollout notes.
Sources / notes:
- OpenAI Sora announcement and technical writeups describing diffusion-transformer, space-time patches, and job-based API workflows.