OpenAI has launched 'Sora', a text-to-video model that allows users to create videos using prompts

OpenAI has launched 'Sora', a text-to-video model that allows users to create videos using prompts.


OpenAI is introducing a new video-generation model named Sora, which the AI company claims can "craft realistic and imaginative scenes based on text prompts." This text-to-video model allows users to generate photorealistic videos up to one minute in length, all from prompts they've written.

Sora is designed to produce "intricate scenes with multiple characters, specific motions, and precise details of the subject and background," as explained in OpenAI's introductory blog post. The model is said to comprehend how objects exist in the physical world and can accurately interpret props, as well as create compelling characters that express lively emotions.

Additionally, the model can generate a video based on a single still image, as well as fill in missing frames in an existing video or extend its length. The demonstrations of Sora included in OpenAI's blog post feature various scenes, such as an aerial view of California during the gold rush and a video simulating a view from inside a Tokyo train. While some of these videos exhibit characteristics typical of AI-generated content—such as anomalies like a moving floor in a museum scene—OpenAI acknowledges that the model "may struggle with accurately simulating the physics of a complex scene." Nonetheless, the results are generally impressive.

A few years ago, text-to-image generators like Midjourney were at the forefront of models' capabilities to transform text into images. However, the advancement in video generation has been remarkable recently, with companies like Runway and Pika showcasing impressive text-to-video models. Google's Lumiere is another significant competitor in this field, offering text-to-video tools and the ability to create videos from a single image.

Currently, Sora is only accessible to "red teamers" who are evaluating the model for potential risks and harms. OpenAI is also providing access to some visual artists, designers, and filmmakers for feedback. The company cautions that while the current model might not accurately simulate the physics of complex scenes or properly interpret certain cause-and-effect scenarios, it is continuously being refined.