OpenAI has officially launched Sora, its advanced artificial intelligence (AI) video generation model, after months of delays. Initially previewed in February, the model has now been made available as a standalone platform, offering premium features exclusively to ChatGPT Plus and Pro subscribers.
Sora is designed to generate high-quality videos up to 20 seconds long, offering resolutions up to 1080p. Users can create videos in widescreen, vertical, and square aspect ratios, as well as upload their own media to extend or remix content. Using text prompts, the AI can generate videos from scratch, while the storyboard interface allows users to set inputs for specific frames.
The model employs diffusion technology, ensuring consistency across frames by analyzing multiple frames simultaneously. It incorporates a transformer architecture and draws on techniques from OpenAI’s DALL-E 3 for enhanced capabilities in video generation.
The platform is currently restricted to ChatGPT Plus and Pro users. Plus subscribers can generate up to 50 videos per month at 480p resolution, while Pro users, at a $200 monthly subscription fee, gain access to 10x more usage, higher resolutions, and longer video durations.
In the case of Sora, the model was trained with a combination of publicly available data and images through creative commons, with OpenAI’s partners such as Shutterstock, Pond5 as well as many other licensed data generated by OpenAI. Additional feedback was also sampled from the employees, red teamers, and trainers to enhance the functionality of the collected data.
Due to ethical issues, OpenAI has placed measures such as the following. Sora has watermarks and metadata this time, too, which the Coalition for Content Provenance and Authenticity (C2PA) has defined. Also, it prevents content that includes abusive contents like deep fakes or obscene content, and also restricts the numbers of videos a user can post to minimized misuse.
This is the new direction OpenAI takes the content creating industry as it launches Sora for the public. Specifically, the provision of advanced video generation technologies to premium users seeks to revolutionize social production of visual artifacts. That being said, as the deployment advances further, Sora’s features and safety measures put it at the forefront of AI video creation, which is still being developed.