Open AI on Thursday unveiled its latest innovation called Sora, a text-to-video model capable of generating one-minute-long videos based on user prompts. While not yet accessible to the public, the announcement sparked a flurry of reactions online.
CEO Sam Altman encouraged prompt ideas on X and demonstrated Sora’s capabilities by generating various videos, including scenes featuring aquatic cyclists, a cooking tutorial, and dogs podcasting on a mountain.
Building on the success of ChatGPT and DALL-E, OpenAI aims to push the boundaries of generative AI with Sora. The model boasts the ability to accurately simulate multiple characters and diverse types of motion, striving to bridge the gap between AI and real-world interaction.
Despite its advancements, Sora still grapples with challenges such as capturing complex physics and spatial details accurately. OpenAI acknowledges that the model may occasionally produce illogical or unnatural results, yet many of the showcased demonstrations exhibit hyper-realistic visual fidelity.
In a move to mitigate potential risks, Sora is currently available to red teamers and a select group of creative professionals for feedback.
OpenAI is also developing labeling tools to distinguish Sora-generated content and implementing safety measures to filter inappropriate prompts.
Furthermore, the company pledges to engage policymakers, educators, and artists worldwide to address concerns and identify positive applications for the technology. By learning from real-world usage, OpenAI aims to ensure the responsible development and deployment of Sora and future AI systems.