OpenAI Launches Sora Turbo for AI Video Generation

OpenAI has officially launched its much-anticipated AI video generation model, Sora, for paid subscribers. The model, introduced as Sora Turbo, allows users to generate up to 20-second-long videos in 1080p resolution. Initially announced earlier this year, its release was delayed for nine months as OpenAI worked to enhance privacy and safety features. Sora is now available as a standalone platform for ChatGPT Plus and ChatGPT Pro users, offering a range of capabilities for creating AI-generated videos from text prompts and storyboards.

Launch Details

• OpenAI’s Sora model, introduced as Sora Turbo, was released for paid subscribers.
• The model can generate AI videos up to 20 seconds long in 1080p resolution.
• Sora is a standalone program accessible through a dedicated website.
• It is available to ChatGPT Plus and ChatGPT Pro users, but new account creations are temporarily restricted.

Features and Capabilities

• Video Generation: Sora uses a diffusion model to anticipate multiple frames at once, ensuring video consistency over the 20-second duration.
• Architecture: The model is based on a transformer architecture, similar to the one used in DALL-E 3, and includes recaptioning techniques.
• Text-to-Video Creation: Sora can create videos from scratch using text prompts and a storyboard interface for precise input for each frame.

Subscription-Based Limitations

ChatGPT Plus Users

• Generate up to 50 videos monthly at 480p resolution.
• Alternatively, create fewer videos at 720p resolution.

ChatGPT Pro Users

• Generate videos in square, widescreen, and vertical aspect ratios.
• Ability to upload videos and images for remixing, blending, and extending content.

Privacy and Safety Features

• Watermarking and Metadata: To prevent misuse of realistic AI-generated content, Sora includes visible watermarks and metadata in line with C2PA standards.
• Content Restrictions: The model blocks the generation of videos that contain sexual deepfakes or child sexual abuse content.
• Data Sourcing: The model’s data comes from public datasets, partnerships with companies like Shutterstock and Pond5, as well as AI trainers and employees.

Privacy and Security Enhancements

• OpenAI has taken significant steps to strengthen the privacy and safety of Sora Turbo, particularly in response to concerns around realistic AI video generation.
• The AI model is designed to prevent harmful uses, such as the creation of deepfake videos, by including protective measures like watermarking and strict content filtering.