Sora 2, OpenAI's latest AI video generation tool, is experiencing significant restrictions that may impact your video creation workflow. This article explains the current limitations and provides guidance on how to work effectively within these constraints.
Sora 2 currently blocks uploads of real people's likenesses unless they've consented, and restricts content involving public figures or minors. Our testing indicates that prompts are being blocked as soon as they contain references to real people, making it extremely difficult to generate videos featuring realistic human subjects.
What Gets Blocked:
Prompts mentioning real people (celebrities, public figures, etc.)
Images uploaded containing real human faces without explicit consent
Content that could be used to create deepfakes or impersonate individuals
Any reference to minors
The platform has inconsistent enforcement of copyright protections, with some copyrighted characters and brands being blocked while others slip through. When attempting to use certain well-known characters like Batman, Mickey Mouse, or Homer Simpson, the system triggers blocks with messages like "This content may violate our guardrails concerning third-party likeness."
However, enforcement appears inconsistent - some branded content may generate successfully while similar requests get blocked. OpenAI has stated they will work with rights holders to block characters at their request and respond to takedown requests.
Be prepared for significant wait times when generating videos with Sora 2:
Sora 2 (Standard):
Standard quality videos: 5-10 minutes generation time
HD quality 15-second videos: 10-15 minutes generation time
According to OpenAI's help documentation, generation may take up to a minute for standard requests, though real-world experience shows longer times
Total expected time: Up to 15 minutes
Sora 2 Pro:
Higher resolution and longer video options
More complex generations with advanced features
Total expected time: Up to 30 minutes
During peak hours, all users, including those on Pro plans, may experience wait times of up to several hours. Video generations are prioritized in queues based on your ChatGPT subscription plan, with ChatGPT Pro users receiving priority.
Tips for Faster Processing:
Generate videos during off-peak hours (typically late night/early morning US time)
Keep video duration shorter (10-15 seconds rather than 20 seconds)
Use standard quality instead of HD when possible for faster results
Given the current restrictions, we recommend focusing your Sora 2 usage on the following types of content:
What Works Well:Object-Only Scenes:
Still life compositions
Architecture and buildings
Product demonstrations
Landscape and nature scenes
Abstract visualizations
Non-Human Subjects:
Animals and wildlife
Animated/cartoon characters (fictional, non-copyrighted)
Illustrated or stylized content
Fantasy creatures
Robots or mechanical objects
Creative Styles:
3D rendered scenes
Paper craft or claymation aesthetics
Painterly or artistic interpretations
Surreal or abstract concepts
What to Avoid:Any prompts mentioning real celebrities, politicians, or public figures
Realistic human faces or people
Well-known copyrighted characters from major entertainment franchises
Recognizable brand logos or trademarked imagery
Content that could be mistaken for real footage of actual people
Instead of "Taylor Swift singing on stage," try "a young woman with long hair performing at a concert venue" (though this may still be blocked due to human likeness restrictions).
Better yet, use: "An animated character performing a concert in a stylized venue, cartoon aesthetic"
The safest approach is to create videos featuring:
"A golden retriever playing in a park at sunset"
"A hummingbird hovering near tropical flowers, macro shot"
"Architectural walkthrough of a modern glass building"
Specify artistic rendering to avoid realism concerns:
"A forest scene in watercolor painting style"
"Clay animation of a kitchen scene with breakfast being prepared"
"Isometric pixel art city street with cars moving"
If you need to include yourself or friends, use Sora 2's official Cameo feature, which requires explicit consent and allows you to control how your likeness is used. This is the only approved way to include real people in your videos.
OpenAI has indicated they are working on giving copyright holders "more granular control over generation of characters" and may adjust these restrictions over time. The current safeguards appear to be precautionary measures as OpenAI navigates the complex legal and ethical landscape of AI-generated video content.
We will update this article as restrictions evolve and OpenAI provides more clarity on what content is permissible.
Real people/likenesses: Currently heavily restricted or blocked in prompts
Brands/copyrighted characters: Inconsistently blocked; avoid to prevent generation failures
Generation time (Sora 2): 5-15 minutes depending on quality settings
Generation time (Sora 2 Pro): Up to 30 minutes for complex, high-quality videos
Best approach: Focus on object-only scenes, animals, nature, architecture, and artistic/illustrated styles
Peak hours: May add several hours of wait time to your generation
If you continue to experience blocked prompts or unexpected restrictions, try:
Removing any references to real people or brands from your prompt
Specifying an artistic or animated style rather than photorealism
Using more generic descriptions of subjects
Generating during off-peak hours for faster processing
For the most current information on Sora 2 restrictions and capabilities, check OpenAI's official documentation and status updates.
