Project Bernini is a research effort focused on developing generative AI for Design and Make industries.
The first experimental Bernini model quickly generates functional 3D shapes from a variety of inputs including 2D images, text, voxels and point clouds.
While Project Bernini is exciting and at the leading edge of generative AI, its models will become increasingly useful and compelling when trained on larger, higher-quality professional datasets.
For the Design and Make industries, accuracy and productivity in 3D modeling is of the utmost importance. Whether you’re an architect, engineer, manufacturer, or filmmaker, the geometric control and accuracy of your digital design is critical to the success of your final product.
That is why Autodesk is obsessed with geometry and our obsession is reflected in our software.
Today, we’re excited to unveil a research effort we’re calling Project Bernini, and its first experimental generative AI model that quickly generates superior 3D shapes from a variety of inputs including a single 2D image, multiple images showing different views of an object, point clouds, voxels, and text. Our first Bernini model is acutely geared toward professional geometric workflows and can generate multiple functional variations of a 3D shape from a given input.
Autodesk is working to create generative models that could be used for different use cases across architecture, product design, entertainment, and more. We’re intently focused on generating functional 3D structures because the articles built or manufactured from the outputs of these models must work in the real world, serving the purpose that the designer has in mind.
A simple example of this would be a water pitcher. Many other 3D generative models might produce shapes that look like a pitcher with textures that improve their superficial appearance in a very specific lighting environment. But the Bernini model generates shape and texture separately and does not confuse or meld those variables. So, water pitchers generated by our model are hollow in the middle and could actually hold water, as any real-world pitcher would need to do.
Just as important, we’re applying a truly generative approach that produces several variants, giving a designer choice and contributing to their creative workflow. According to Autodesk’s 2024 State of Design & Make report, 78% of business leaders believe AI will enhance their industry and 79% agree that AI will make their industry more creative.
The Autodesk AI Lab within Autodesk Research is a diverse team of scientists and specialists working on applying AI to aid in difficult tasks in design, manufacturing, architecture, engineering, and construction, as well as the creative processes in media and entertainment. They trained the Bernini model on ten million diverse 3D shapes—a composite dataset made up of publicly available data, a mixture of CAD objects and organic shapes. Research that contributed toward this work was published earlier this year by the AI Lab in collaboration with the Chinese University of Hong Kong.
At this stage, Project Bernini is strictly experimental and is not available for public use.
As the team works to refine this technology, Autodesk is actively engaging potential partners to further study and improve the performance of generative models for design and make. While Project Bernini is exciting and at the leading edge of generative AI, its models will become increasingly useful and compelling when trained on larger, higher-quality professional datasets. The Bernini model is specifically structured to offer efficient training and fine-tuning to easily allow this.
If trained on buildings, the models could generate geometrically rigorous creative designs and inspire a new generation of buildings and architects. If trained on video game character models or fantasy environments, they could produce fascinating new creatures or virtual worlds. If trained on car designs, they could assist in imagining an innovative new series of vehicles.
