News

Nvidia brings generative AI to 3D models in TurboSquid

Shutterstock’s 3D model generator is trained on licensed, ethically-sourced data.

Jon Peddie

Nvidia and Shutterstock have partnered to introduce a groundbreaking generative AI tool for creating 3D models, announced at Siggraph 2024. This tool, integrated into Shutterstock’s TurboSquid library, allows users to generate 3D assets using text and image prompts. It offers rapid previews, high-quality conversions with PBR materials, and 360-degree HDRI background generation. The technology, trained on ethically sourced data, is initially available to enterprise customers and integrates with software like Blender.

TurboSquid
(Source: TurboSquid)

Numerous generative AI tools exist for creating images, such as Stable Diffusion and Microsoft’s recently introduced Designer. Nvidia, however, is taking generative AI to a new level. At Siggraph 2024, the company announced a collaboration with Shutterstock that enables users to generate 3D models using generative AI.

While there are some AI tools for 3D model generation such as Meshy and Vast, Shutterstock’s offering stands out as the first tool introduced so far for Nvidia’s Omniverse platform. The models will be available in Shutterstock’s TurboSquid library, which currently hosts a variety of 3D assets. Users can create new models by providing text and image prompts, and they can edit these assets in other applications, with exports available in various popular file formats.

Nvidia states that designers can preview their generated assets in as little as 10 seconds. After reviewing the preview, they can choose to convert it into a high-quality 3D model, applying physically-based rendering (PBR) materials in the process. Additionally, users will be able to create 360-degree HDRI backgrounds to enhance lighting effects in virtual scenes.

Nvidia emphasizes that Shutterstock’s 3D model generator is trained on licensed, ethically sourced data, ensuring the assets are safe for commercial use. However, as with other generative AI tools, the origins of the dataset can be difficult to ascertain.

In addition to the partnership with Shutterstock, Nvidia says it is enhancing its generative AI features with Getty Images, claiming the service can now produce images twice as fast and with improved quality.

Shutterstock’s 3D feature is launching in beta, initially available only to enterprise customers, with no current plans to extend access to non-commercial users. At Siggraph, the company demonstrated how the tool integrates with Blender, allowing for the direct generation of 3D assets within the application. HP is also showcasing physical prototypes of 3D models created using Shutterstock’s technology.