MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) is helping define a new era of 3D printing, one where printed parts can feel like fabric, move like muscles, and carry embedded electronics. These aren’t isolated breakthroughs, but signs of a broader effort to reshape how we design, fabricate, and interact with printed objects.
Over the past few years, CSAIL researchers have steadily rolled out innovations that expand what 3D printing can do. These projects blend artificial intelligence (AI), materials science, and automation to make additive manufacturing (AM) smarter, more intuitive, and more useful across industries.
Making 3D Models Touchable: TactStyle
Traditional 3D modeling focuses on how things look—the shape, color, and surface details—but usually doesn’t consider how they feel to the touch. CSAIL’s new system, TactStyle, changes that. It lets users design both the appearance and texture of an object just by uploading an image. For example, you can make a 3D print that not only looks like wood but also feels like wood grain.
Developed by PhD student Faraz Faruqi and Associate Professor Stefanie Mueller, TactStyle was introduced in March 2025 and presented at this year’s Conference on Human Factors in Computing Systems (CHI2025) in Yokohama, Japan.
Faraz Faruqi at MIT CSAIL. Image courtesy of Mike Grimmett/MIT CSAIL.
TactStyle separates visual stylization (what something looks like) from geometric stylization (how it physically feels), allowing designers to replicate real-world textures more accurately. It’s especially useful in areas like home decor, where realism is important, or in education, where students with visual impairments can benefit from tactile learning aids.
The system can capture an image of a woven basket and reproduce not only its appearance, but also its texture, resulting in prints that combine both visual and tactile realism. It marks a step toward a more sensory-aware approach to 3D modeling.
Printing Motion: The Xstrings Method
If TactStyle adds a sense of touch to 3D design, Xstrings brings in movement, allowing printed objects to curl, twist, or grip right off the print bed.
Objects that move, like robotic limbs, flexible sculptures, or wearable tech, usually need lots of separate parts and careful manual assembly. But CSAIL’s Xstrings method changes that. Released in early 2025, Xstrings streamlines the process by combining design software and 3D printing into a single system.
Xstrings is a hybrid design and fabrication tool that embeds built-in motion systems directly into 3D printed objects. These tensioned cables can drive actions like bending, twisting, curling, or gripping—all printed in one go, with no screws, motors, or extra assembly needed.
Examples of Xstrings. Image courtesy of MIT CSAIL.
The research team, led by MIT CSAIL postdoctoral researcher Jiaji Li and senior author Stefanie Mueller, demonstrated the method by printing everything from a robotic lizard and a wall sculpture with peacock-like motion to a claw that can close into a fist. Users can choose different types of motion (like coiling or compressing), and even combine them, allowing for both simple and complex movements controlled by pulling a string.
The system runs on common fused deposition modeling (FDM) 3D printers, which carefully print around the cables. Tests showed that the embedded strings could withstand more than 60,000 pulls before breaking. According to the researchers, Xstrings can reduce production time by about 40% compared to manually assembling cable-driven devices.
It has potential use in soft robotics, interactive art, and even adjustable clothing. Looking ahead, the team believes Xstrings could be used in places with limited space and tools, like space stations, making it possible to print functional, moving robots on demand.
How Xstrings is 3D printed. Image courtesy of MIT CSAIL.
Toward Fully 3D Printed Electronics
One of the most exciting and challenging frontiers in 3D printing is the creation of active electronics, or devices that can store, process, or regulate electrical signals. Typically, these components rely on silicon semiconductors and must be built in specialized clean rooms. But in late 2024, a team at MIT took a major step forward by demonstrating the first fully 3D printed, semiconductor-free logic gates, the kind used in basic computation.
Led by Luis Fernando Velásquez-García, a principal research scientist at MIT’s Microsystems Technology Laboratories, and graduate student Jorge Cañada, the team created resettable fuses using a copper-infused biodegradable polymer. These devices were printed in a single step using off-the-shelf extrusion 3D printers. While they aren’t as powerful as silicon transistors yet, they’re a big step forward because they can still handle basic control functions, like turning a motor on and off.
The devices are made from 3D printed traces of the copper-doped polymer. Image courtesy of Luis Fernando Velásquez-García/MIT CSAIL.
The research, published in Virtual and Physical Prototyping, shows how on-demand, low-cost, and low-waste fabrication of electronics can become more accessible to labs, businesses, and even hobbyists, particularly during times of supply chain disruption. It hints at a future where electronics can be manufactured locally, as needed.
Sustainable and Multi-Textured Printing: Speed-Modulated Ironing
While some projects focus on what printed objects can do, others look at how we can make printing itself more sustainable, without giving up aesthetics or functionality.
Making 3D printing more sustainable often means finding ways to do more with less, especially using just one material to create complex results. In October 2024, MIT CSAIL and Delft University of Technology (TU Delft) introduced a method called speed-modulated ironing that helps do exactly that.
The technique uses a dual-nozzle 3D printer. One nozzle lays down the filament, and a second heated nozzle moves over the surface afterward, adjusting its speed to change how the object looks and feels. Faster passes create a glossy surface, while slower passes result in a matte or textured finish.
This process lets designers create multi-textured, multi-toned objects without switching materials or doing extra post-processing. It’s especially useful with filaments like wood-filled or temperature-sensitive plastics. The team even showed how it can be used to embed things like QR codes directly into prints using texture alone.
Speed-modulated ironing is more than just a visual trick; it’s a step toward greener, more efficient 3D printing. By using a single filament and avoiding wasteful steps, it reduces material use and energy.
The project was presented at the ACM UIST 2024 conference and included researchers from both MIT and TU Delft, including Mehmet Ozdemir, Marwa AlAlawi, Mustafa Doga Dogan, Stefanie Mueller, and Zjenja Doubrovski.
3D Models with AI: Style2Fab
Customizing a 3D model is usually something only trained designers with complex CAD software can do. CSAIL’s Style2Fab system is changing that, using AI to make customization easier for everyone.
Released in 2023, Style2Fab is designed for makers of all skill levels, especially those working with open-source models. It lets users modify 3D models using text prompts. For example, typing “make this vase look like a seashell” transforms the model’s appearance while keeping its function intact.
MIT’s Style2Fab. Image courtesy of MIT CSAIL.
This system was also developed by Faruqi, CSAIL professor Stefanie Mueller, and collaborators, including Megan Hofmann at Northeastern University. Their work was detailed in the paper, Style2Fab: Functionality-Aware Segmentation for Fabricating Personalized 3D Models with Generative AI.
Style2Fab’s key feature is that it doesn’t just decorate the surface; it also enhances the interior. It uses deep learning to analyze and divide a model into functional and aesthetic parts, allowing users to safely modify colors, textures, or shapes without changing its performance. For example, someone could redesign a 3D printable wrist splint to match their clothing, without affecting its support function.
By removing the need for expensive tools or design expertise, Style2Fab helps democratize digital fabrication. It’s especially promising for DIY assistive devices, where both functionality and personal style matter. As Faruqi puts it, the tool is meant to “make 3D printing easier to experiment with and learn from, whether you’re a beginner or an expert.”
Vision Systems and Real-Time Monitoring
Bringing all these innovations together requires more intelligent systems. That’s why CSAIL is also rethinking the printers themselves. In a recent collaboration with ETH Zurich and MIT spinout Inkbit, they developed a vision-based 3D printer that can monitor the printing process in real-time and correct mistakes as they print.
This printer uses computer vision and feedback loops to monitor each printed layer, detect defects, and adjust the next layer accordingly. It allows the system to handle materials that usually don’t work with traditional methods, such as slow-curing resins or components that require tight dimensional control.
By combining real-time sensing with machine learning, this approach offers a more intelligent way to manufacture, one that can self-correct, adapt to different materials, and deliver precision at scale.
MIT Vision Jetting. Image courtesy of MIT.
Taken together, these projects reflect MIT CSAIL’s unique approach to 3D printing. It’s not just about printing faster or prettier objects. It’s about building a system where hardware, software, materials, and intelligence work together to create tools that are functional, intuitive, and impactful.
From electronics to education, from art to accessibility, CSAIL’s researchers are pushing AM in new and unexpected directions. They’re laying the groundwork for what 3D printing may soon become: interactive, personalized, and increasingly embedded in how we design and build.
