The popularity of AI image generation has led us to the next frontier: video generation. Nearly every major player in the AI space, including Google and OpenAI, is developing text-to-video generators. Stability AI just released its latest AI video model, but its take is a bit different from the rest.
Also: This AI video platform will assemble a short for you from start to finish
On Wednesday, Stability AI unveiled Stable Video 4D, an AI model that can transform a single object video into multiple novel-view videos of the object from different angles and perspectives. This allows users to generate realistic, multi-angle videos from just one video, as seen below.
When using the model, a user would start by uploading a single video and delineating what 3D camera poses they'd like to see rendered. Then, in 40 seconds, Stable Video 4D generates five frames across eight views, with the entire 4D optimization taking 20 to 25 minutes, according to the blog post.
Stability AI sees Stable Video 4D as a tool capable of helping working professionals across various industries in the near future.
"Our team envisions future applications in game development, video editing, and virtual reality," said Stability AI. "Professionals in these fields can significantly benefit from the ability to visualize objects from multiple perspectives, enhancing the realism and immersion of their products."
Even though Stability AI has released other AI video generators, such as Stable Video Diffusion, which converts images into videos, and Stable Video 3D, which generates 3D objects from single images, this is the company's first video-to-video generator, which the company calls an "exciting milestone."
Also: The best AI image generators: Tested and reviewed
Alongside the AI model, the company also released a technical report showcasing the model's development, including the methodologies, challenges, and more.
Stable Video 4D is currently available on Hugging Face. Stability AI says it is actively working on improving the model.