OpenAI Introduces Sora: An AI Video Generation Tool
OpenAI has launched Sora, its latest AI tool capable of generating videos from text prompts or images. While the tool offers impressive capabilities, OpenAI is cautiously withholding a key feature: the ability to generate videos using uploaded photos or videos of real people as a reference. This decision stems from concerns about potential misuse, including deepfakes and misinformation. Learn more about Sora's release.
Safety and Moderation Measures
OpenAI emphasizes its commitment to responsible AI development. The company is implementing various safety measures, including stricter moderation for videos potentially depicting minors and metadata to identify Sora-generated content. Similar AI image generators have faced similar challenges. OpenAI acknowledges the ongoing challenge of balancing creative expression with preventing misuse and encourages user feedback on its moderation efforts.
Copyright and Training Data
To address copyright concerns, OpenAI is using "prompt re-writing" to prevent Sora from generating videos that mimic the styles of living artists. The company states that Sora was trained on a mix of publicly available data, proprietary data, and custom datasets. This approach aims to mitigate potential legal issues related to copyright infringement, a concern raised in cases involving other AI development projects.
Features and Availability
Sora offers a range of features, including the ability to create multiple video variations, edit existing videos, and create storyboards. The tool is currently available to ChatGPT Plus and Pro subscribers in select countries.