A new generative AI tool called MechStyle, developed by researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), enables users to create personalized 3D-printed objects that are both stylish and functional. The system allows users to input a base 3D model—such as a wall hook—and a text prompt like “generate a cactus-like hook.” Using AI, the tool transforms the model’s geometry to match the described style while preserving its structural integrity. Previously, AI-driven 3D stylization often led to impractical designs. A prior study by the team found that only about 26 percent of modified 3D models remained structurally viable, as the AI lacked understanding of physical forces like stress and load distribution. MechStyle addresses this by integrating finite element analysis (FEA) simulation with adaptive scheduling. This allows the system to predict how design changes will affect strength and stability in real-world use. The system works by identifying weak points in a model during stylization and adjusting the AI’s modifications to prevent structural failure. It can either halt changes when stress thresholds are approached or make gradual refinements to maintain durability. Testing with 30 different models—styled to resemble bricks, stones, cacti, and more—showed that MechStyle could produce objects with up to 100 percent structural viability. MechStyle offers two modes: a freestyle mode for quick visual exploration of different styles, and a MechStyle mode that rigorously evaluates structural impact. Users can experiment with creative ideas and then test them for real-world usability, making the design process faster and more accessible. The tool can create a range of personalized items, from a pillbox with a rocky, checkered texture to a lampshade resembling red magma, or a pair of glasses with fish-scale patterns. It also shows promise in assistive technology, such as custom finger splints or utensil grips tailored to individual needs. While MechStyle currently cannot repair structurally unsound models, the team plans to expand its capabilities to fix flawed designs in the future. They also aim to evolve the system to generate 3D models from scratch based on user prompts, rather than relying on pre-existing templates. This would allow anyone, even those without 3D design experience, to create unique, printable items like a custom bowl. The research was led by Faruqi, a CSAIL researcher and lead author, alongside senior author Stefanie Mueller, and colleagues Leandra Tejedor, Jiaji Li, Amira Abdel-Rahman, Martin Nisser, Vrushank Phadnis, Varun Jampani, Neil Gershenfeld, and Megan Hofmann. The work was supported by the MIT-Google Program for Computing Innovation and presented at the Association for Computing Machinery’s Symposium on Computational Fabrication in November. According to Google Research Scientist Fabian Manhardt, who was not involved in the study, MechStyle is a major step forward in 3D style transfer, solving a key challenge in making AI-generated 3D designs both creative and usable.