Leveraging State-of-the-Art Generative AI Models in Autodesk Maya

The past few months have seen a rapidly evolving landscape of new Generative AI tools. From Large Language Models like GPT-4, to image diffusion techniques for generating detailed images from text, the applications of Generative AI continue to expand. As these models evolve, they are finding their way into more aspects of digital content creation, including 3D modeling and animation.

In this video tutorial, I delve into how we can leverage such models from within Autodesk Maya, an industry-standard 3D animation and modeling software used heavily in the film, TV, and games industries. I demonstrate how to use Maya’s Python API with OpenAI’s Point-e to generate 3D shapes as point clouds from complex text prompts. I also show how to visualize the point cloud in Maya’s viewport with Bifrost, and present a few practical techniques for manipulating the point clouds in Bifrost as part of a typical artist workflow.

By making state-of-the-art, generative AI models like Point-e accessible to artists from within their familiar design environment, they can begin to incorporate these new capabilities seamlessly into their artistic workflows.

Evan Atherton is a Principal Research Scientist at Autodesk.