On Monday, OpenAI launched Sora, a revolutionary AI model that generates videos from text descriptions. The new model, named Sora Turbo, is now accessible to paid ChatGPT users in the US and “most other countries” via Sora.com.
Video Generation and Pricing
Sora Turbo allows users to create videos ranging from 5 to 20 seconds in length, supporting various aspect ratios and resolutions between 480p and 1080p. However, generating these videos requires credits, the cost of which varies:
- 480p video: 20 to 150 credits
- 720p video: 30 to 540 credits
- 1080p video: 100 to 2,000 credits
OpenAI has not clarified all factors influencing these costs.
ChatGPT Plus subscribers, paying $20 per month, receive 1,000 credits, enabling up to 50 priority videos in 720p at 5 seconds each. Pro plan users, paying $200 per month, get 10,000 credits, allowing 500 priority videos in 1080p at 20 seconds each. Pro users also gain access to unlimited low-priority video generations, up to five simultaneous generations, and watermark-free downloads. By default, Sora videos include visible watermarks and C2PA metadata, marking them as AI-generated.
Advanced Tools and Features
Sora offers several tools for video creation and editing:
- Re-mix Tool: Edits existing videos.
- Storyboard Interface: Creates videos from sequential prompts.
- Blend Tool: Combines two videos while retaining key elements.
- Loop and Re-cut Options: Enable further customization.
Users can generate multiple versions of clips based on prompts or images, providing flexibility for creative projects.
Challenges and Limitations
Despite its potential, Sora faces technical limitations. Marcus Brownlee, a prominent video blogger known as MKBHD, tested an early version and noted that generating a 1080p video took a few minutes. He observed issues common to AI video generators, including a lack of object persistence.
For example, objects in Sora videos may move inconsistently or disappear and reappear unpredictably. Additionally, animations involving legs often exhibit errors, with front and back legs switching positions or otherwise behaving unnaturally.
Safeguards and Ethical Use
OpenAI has implemented robust safeguards to prevent misuse of Sora. The model will not generate videos featuring individuals under 18, violent content, explicit themes, or violations of copyright. Sora also avoids creating videos based on images of public figures, logos, or recognizable characters. Misuse of uploaded media can result in account suspension or bans, adds NIX Solutions.
Applications and Future Developments
Sora shows promise for generating styled intros, animations, abstracts, and still frames, though it is less suitable for creating photorealistic videos. OpenAI acknowledges that Sora is an early version and expects users to encounter bugs.
“It’s not perfect, but it’s at a stage where we think it’s going to be really useful for augmenting human creativity,” said Will Peebles, a technical lead on Sora.
For those without a ChatGPT subscription, OpenAI offers a feed to view AI-generated videos created by other users. While Sora is available in the US and many other countries, OpenAI CEO Sam Altman indicated that deployment across Europe and the UK might require additional time. Yet we’ll keep you updated as more integrations become available.
Looking Ahead
Sora’s debut marks a significant step in AI-powered video creation, offering exciting opportunities for creative professionals and hobbyists alike. With ongoing updates and improvements, OpenAI hopes to unlock new possibilities in video production.