OpenAI releases Sora: All you need to know about AI video generation model

With Sora, OpenAI enables ChatGPT Pro subscription members to create 20-second videos in up to 1080p resolution, with watermark-free downloads available

Microsoft-backed artificial intelligence company OpenAI has announced the rollout of its video generation model Sora, which is designed to take text, image, and video inputs and generate a new video as an output. It is a standalone service and not integrated into ChatGPT. The video generation tool is accessible through its own dedicated website and operates separately from ChatGPT. However, it requires a ChatGPT Plus or the new Pro subscription.

 

OpenAI has also introduced a faster version, Sora Turbo, which the company claims significantly outperforms the original Sora model previewed in February this year.

 

What is new in OpenAI’s Sora?

 

OpenAI stated that premium subscribers can generate videos in 1080p resolution, up to 20 seconds long, and in widescreen, vertical, or square aspect ratios. Sora accepts text, images, and videos as input prompts, enabling users to extend, remix, and blend content or generate entirely new creations from text.

The updated Sora interface includes a storyboard tool, allowing users to specify inputs for each frame with precision. The interface also features sections like "Featured" and "Recent" feeds, showcasing user creations, along with “All Videos,” “Favourites,” and “Upload” sections for easier navigation.

 

Availability and subscriptions

 

The Sora video generator is available with ChatGPT Plus and the new ChatGPT Pro subscriptions. With ChatGPT Plus, users can generate videos up to 5 seconds long in 720p resolution, with a limit of 50 videos.

 

The Pro subscription offers enhanced capabilities, including 20-second videos in 1080p resolution, five concurrent generations, and a usage limit of 500 video creations. Additionally, Pro subscribers can download videos without watermarks.

 

Limitations of OpenAI’s Sora

 

OpenAI has implemented safeguards, including default visible watermarks on videos and an internal search tool to verify content origins using technical attributes.

 

To prevent misuse, particularly with materials like deepfakes, OpenAI is currently restricting uploads of content featuring people. However, the company plans to enable this feature in the future after refining deepfake mitigation measures.