OpenAI unveils Sora, a new AI model for generating photorealistic videos from text prompts

OpenAI has introduced its new Text-to-Video AI model, Sora, capable of creating highly realistic video scenes from text prompts. Sora can generate photorealistic videos of up to a minute in length, featuring complex scenes with multiple characters and specific movements based on user-written prompts. The model is versatile, capable of generating videos from still images, filling in missing frames in existing videos, or extending them.

Sora is actually quite impressive, it demonstrates an understanding of how objects exist in the physical world, can interpret props accurately, and create characters with vibrant emotions. Despite this, the model may show signs of AI-generated content, including inaccuracies in simulating complex physics within scenes, and interpreting cause and effect scenarios.

The model is currently only accessible to "red teamers" evaluating the potential risks and harms of the model, as well as a select group of visual artists, designers, and filmmakers for feedback, although it is expected to be available to the general public soon. In the meantime, some of the most recognizable AI Video Generation alternatives are Runway, Stable Video Diffusion, and Pika Labs, which, similar to Sora, have shown impressive text-to-video tools and allow users to create videos from a still image.