Generating 3D images used to be an extremely difficult task involving complex wireframes, software, and hardware, but things have changed dramatically.Stability AI A new generative AI technology called Stable Fast3D was recently announced that can quickly generate 3D images from a single image.
The most amazing thing is that according to Stability What AI Says, the new model can generate 3D images in half a second.This processing speed is a qualitative leap compared to the previous model, which might take several minutes to generate similar results.Stable Fast3D completes the same task 1,200 times faster than before.
As early as March, Stability AI released Stable Video3D (SV3D), which takes 10 minutes to generate 3D assets. Today's Stable Fast3D is a huge improvement.
Stability AI expects this new model to be useful in a variety of industries, including design, architecture, retail, virtual reality, and game development. The model is available through the Stable Assistant chatbot, the Stability AI API, and the community license Hugging Face.
Stable Fast3D principle
Stable Fast3D was not built from scratch, but evolved from the previous TripoSR model. In March, Stability AI partnered with 3D modeling provider Trip AI to create fast 3D asset generation technology.
In the research paper, the researchers detailed the innovative working method of Stable Fast3D. The core of Stable Fast3D is to use an enhanced transformer network to generate a high-resolution three-plane, 3D volume representation from the input image. The network is designed to efficiently handle larger resolutions without significantly increasing computational complexity, thereby achieving finer detail capture and reducing aliasing artifacts.
The researchers also detailed an innovative approach to material and lighting estimation. The material estimation network predicts global metallic and roughness values using a novel probabilistic approach that has the potential to improve image quality and consistency.
Stability AI is perhaps best known for its Stable Diffusion text-to-image generation technology, but it has been working on 3D since at least November 2023, and its Stable Video3D in March of this year improved the quality of 3D image generation and viewing experience. Not only that, last week the company also announced Stable Video4D, which adds a time dimension to short 3D video generation.