Sora — Video Generation
Sora is OpenAI's text-to-video and image-to-video model. It accepts a text description (and optionally a source image) and generates a short video clip. With Sora 2, OpenAI moved from an experimental research demo to a production-grade video generation tool capable of 1080p full HD output with physically accurate motion.
Sora 1 (December 2024)
Sora was publicly launched in December 2024, having been previewed earlier that year. The initial version produced clips up to approximately 20 seconds in duration and demonstrated impressive visual quality for an early release, but had limited user control over camera angles, pacing, and world continuity. Physics behaviour was sometimes inconsistent — objects could behave in ways that violated real-world expectations. Resolution and generation options were constrained.
Sora 2 (September 30, 2025)
Sora 2 was a major generational upgrade that addressed most of the shortcomings of the original. The headline improvements:
Technical Capabilities
- Resolution: 1080p Full HD as the standard output
- Duration: 15–25 seconds per clip
- Audio: Synchronized dialogue and sound effects
- Physics: Accurate motion — objects, liquids, and fabrics behave realistically
Control and Consistency
- Multi-shot instructions: Direct camera movements, cuts, and scene transitions
- World state persistence: Characters and objects remain consistent across cuts
- More controllable: Better instruction following for detailed creative direction
Plan Limits
| Plan | Resolution | Duration | Generations | Model Access |
|---|---|---|---|---|
| Free | 720p | Up to 5 seconds | Limited per month | Sora 1 |
| Plus | Up to 720p | Longer | ~50 videos at 480p or fewer at 720p/month | Sora 1 / Sora 2 |
| Pro | Up to 1080p | 15–25 seconds | 10× Plus usage | Sora 2 Pro |
Pro plan subscribers get access to Sora 2 Pro — the highest-quality variant — along with 10× more generation credits than Plus, higher resolutions, and longer clip durations. Sora generation is included in the Pro plan subscription rather than priced separately for ChatGPT users.
Availability
Sora is available in ChatGPT on iOS (iOS 18 or later) and via the web interface. Geographic availability was initially limited to the US and Canada. An API for Sora was described as "coming soon" as of October 2025 but had not launched at that time. Check platform.openai.com for current API availability.
Use Cases
- Product demos: Short product walkthroughs or explainer clips generated from a script and reference image
- Social content: Animated visuals, lifestyle clips, or motion graphics for social media
- Storyboarding: Rough visualisations of scenes before committing to production filming
- Creative exploration: Rapid iteration on visual concepts for directors, designers, and marketing teams
Current Limitations
Important Limitations
- Policy enforcement: Sora includes safety systems that reject requests for deepfakes, non-consensual intimate imagery, and content designed for misinformation. Policy is actively enforced.
- Complex physics: Highly intricate scenes (dense crowd interactions, complex fluid dynamics at small scale) may still produce artefacts.
- No API yet: As of late 2025, Sora is only accessible via ChatGPT — not programmatically via the API.
- Geographic restrictions: Not available in all countries at launch.
Checklist
- What are the headline output specifications of Sora 2 in terms of resolution and duration?
- What does "world state persistence across cuts" mean for video generation quality?
- Which plan provides access to Sora 2 Pro with the highest generation limits?
- What type of content does Sora's safety system actively reject?
- What input types does Sora accept — is it text-only?