OpenAI recently introduced its new video generation tool, Sora, which allows users to create videos from text, images, and existing footage. However, one of its most intriguing features—generating videos using photos or clips of real people—has been restricted to a select group of users for now. OpenAI aims to refine its safety measures before making this feature widely available.
This feature enables Sora to use an uploaded photo or video as a "seed" to create new video content. While OpenAI acknowledges the creative potential of this capability, the company has expressed concerns about its misuse, such as creating deepfakes or other harmful content.
“The ability to generate videos using real people as references carries risks, so we’re taking an incremental approach to learn from early usage patterns,” OpenAI stated in a blog post. To address these risks, the company has temporarily restricted these videos from being shared on Sora’s discovery feed, which showcases user-generated content.
Rohan Sahai, Sora’s product lead, emphasized OpenAI’s cautious rollout during a livestream presentation. He noted the company’s dual focus on preventing illegal activity while supporting creative expression.
Generative video tools like Sora have sparked debate due to concerns about deepfakes and misinformation. A report from ID verification service Sumsub highlighted that deepfake fraud increased more than tenfold between 2022 and 2023.
To mitigate these risks, Sora uses a filtering system to detect and moderate content depicting individuals under 18. OpenAI also applies stricter content moderation for potentially harmful content, such as depictions of violence or self-harm.
OpenAI has added metadata to all Sora-generated videos that comply with the C2PA technical standard, enabling platforms to identify their origin. While users can remove the metadata, OpenAI promotes it as a way to ensure transparency and accountability.
To avoid copyright issues, Sora also includes a “prompt re-writing” mechanism. This feature prevents users from generating videos in the style of living artists without permission. OpenAI acknowledges the ongoing legal disputes over AI training data but asserts its efforts to respect creator rights.
Sora offers several creative tools for video generation. Users can create multiple variations of a video clip based on a text prompt, remix existing videos, and combine elements from two videos using a Blend tool. Additional features like Storyboard and Re-cut allow users to sequence and fine-tune their videos.
For now, Sora is available to subscribers of OpenAI’s ChatGPT Pro and Plus plans in select countries. OpenAI remains committed to improving the tool while balancing innovation with responsible use.
This expository rewrite retains the original facts while presenting them in a clear and accessible manner.
