OpenAI Sora Launched For Premium ChatGPT Users, Can Generate Up To 20-Second-Long AI Videos

OpenAI has finally released Sora Turbo, a video-generation AI model that can generate up to 20-second-long videos with 1080p resolution.

Hyderabad:OpenAI, a leading AI research company, finally launched its much-awaited AI video generation model Sora on Monday for its paid subscribers. The released version is introduced as Sora Turbo, which can generate videos for up to 20 seconds long in 1080p resolution. It is released as a separate program available at a dedicated website and usable with specified rate limits.

This year in February, the AI model was unveiled but was repeatedly delayed as the company was strengthening the privacy and safety parameters of the model. After a delay of nine months, OpenAI has finally launched Sora as a standalone platform. It is currently available to ChatGPT Plus and ChatGPT Pro subscribers. Notably, OpenAI has restricted new account creations on the platform for the time being.

The OpenAI Sora is a diffusion model that can anticipate many frames at a time to maintain the consistency of the video over 20 seconds. It uses a transformer architecture-- a neural network that uses multiple blocks to process data entered by the user and takes recaptioning technique from DALL-E 3. Moreover, this AI model can create videos from scratch using text prompts and a storyboard interface helps users to place important inputs for each frame.

ChatGPT Plus users can generate 50 videos at 480p resolution or create fewer videos at 720p resolution every month. Meanwhile, ChatGPT Pro users can generate videos in square, widescreen, and vertical aspect ratios. Furthermore, Pro users can upload their videos and images to remix, blend, and extend the content into a generated video.

OpenAI claims that it sourced model data from public datasets, data partnerships, and commissioned sources like Shutterstock Pond5, as well as from AI trainers, red teamers, and employees.

To make Sora Turbo safe from misuse associated with realistic AI video generation models, OpenAI has added metadata and visible watermarks as per the standards set by the C2PA (Coalition for Content Provenance and Authenticity). The company also claims to protect media uploads that include people and will block the generation of videos that have any form of sexual deepfakes or child sexual abuse.