This is 3D Gen, Meta’s new AI that creates high-quality 3D textured models in less than a minute

Meta has presented its new Artificial Intelligence (AI) model 3D Gen,which offers capabilities for create textured 3D models from text cues,with results that present high-quality shapes and textures in less than a minute.

The technology company led by Mark Zuckerberg continues to work to explore new AI capabilities, not only when it comes to generating images or video clips, but also to create or improve 3D modeled content.

In this framework, the Meta research team has shown a New system for creating or retexturing 3D objects from text prompts,which has been called 3D Gen, and which is capable of offering results with a speed between 3 and 10 times higher, equivalent to less than a minute.

As explained by Meta in a statement, through a new and rapid process, 3D Gen offers the creation of 3D assets with “instant high fidelity” of quality shapes and textures.

The realism of 3D Gen is due, in part, to the fact that this technology supports physically based rendering (PBR), allowing relighting of 3D assets in real-world applications. It also supports generative retexturing of previously generated or artist-created 3D shapes. All this, using text inputs provided by the user.

The company has also stressed that the technology works with two key technical components, Meta 3D AssetGen and Meta 3D TextureGen,which have also been developed by the Meta research team for the Text to 3D and Text to Texture Generation,respectively.

By combining both tools, 3D Gen represents 3D objects in the display space, in the volumetric space and in the UV space, i.e. texture space. This achieves a 68 percent success rate over single-stage models,according to the technology company.

Meta has pointed out that, after comparing 3D Gen with “numerous industry baselines”, it is a technology that surpasses in terms of indication fidelity and visual quality, as well as being “significantly faster”.