What happens when the cutting edge of AI meets the intricate world of parametric design? Kedar Undale's 'Bendo Bricks' project dives headfirst into this fascinating intersection, exploring a future where artificial intelligence doesn't just assist, but actively collaborates in shaping our built environments.
Imagine a scenario where complex architectural forms, often a headache for designers to visualize through traditional rendering methods, could be brought to life with unprecedented speed. Generative AI tools promise this efficiency, but they often falter when faced with the unique, bespoke geometries of parametric architecture. These custom forms frequently fall outside the typical data sets these AI models are trained on, leading to results that can be a bit… off. Think visually inconsistent or geometrically inaccurate outputs. It’s like asking an artist who’s only ever painted landscapes to suddenly create a hyper-realistic portrait – they might get the general idea, but the fine details could be lost.
But here's where it gets truly innovative: Kedar Undale isn't looking to AI as a magic wand to replace human design thinking. Instead, he’s positioning it as a collaborator. This isn't about a shortcut; it's about a partnership where AI works within the carefully defined boundaries of geometry, material logic, and the architect's original intent. This philosophy is the bedrock of his "AI and Parametric Negotiation" workshops, where designers learn to build precise systems and then allow for controlled, probabilistic variations to emerge. It’s a dance between human intention and algorithmic exploration.
And this is the part most people miss: The latest iteration of these workshops, currently available through PAACADEMY, builds upon these foundational ideas with cutting-edge workflows. We're talking about advanced image editing pipelines and single-image refinement processes, much like those showcased in the 'Bendo Bricks' project itself. These aren't just minor tweaks; they represent a significant leap in how AI can interact with design.
A pivotal moment in this workflow occurred around 2025 with the advent of sophisticated image editing AI models. While we've seen these tools used for straightforward tasks like changing a daytime scene to nighttime or altering the seasons, newer models, such as Qwen Image Edit and Gemini 3 Pro Image, have demonstrated a remarkable ability to generate detailed design refinements. All it takes is a clear prompt and a base image – a powerful combination for iterative design.
The 'Bendo Bricks' project itself unfolded in two distinct stages:
Stage 1: Crafting the Parametric Framework
This initial phase was all about meticulously developing the form. Using Grasshopper 3D within Rhino3D, Undale began by generating a single, doubly curved surface. This foundational surface was then populated with bricks, orchestrated by a parametric system. To inject complexity and consider performance, specific bricks were strategically placed based on the curvature of the façade. This thoughtful dispersion created a jali-like pattern, a beautiful architectural element that allows for controlled light to filter into the interior spaces. To maintain a balance between openness and enclosure, fixed glass panels were seamlessly integrated into these openings.
To ensure the design logic was crystal clear, a single animation sequence was exported from Grasshopper. This visual narrative clearly illustrated the evolution of the geometry, transforming from a simple surface into a structured, brick-clad façade.
Stage 2: Embracing Probabilistic Interpretation
The second stage brought generative image editing into the mix. The primary hurdle here was establishing an accurate base render. This was masterfully achieved using ComfyUI, a node-based interface that allowed for the integration of powerful tools like ControlNet and IPAdapter. These components were crucial for preserving the exact geometry of the parametric model while simultaneously allowing for stylistic interpretation. After numerous iterations, a stable and accurate base image was produced. From this solid foundation, the image editing workflows were applied. The core of this process involved a single-image pipeline within ComfyUI. Here, the uploaded render was meticulously analyzed and enhanced, all while remaining remarkably true to the original geometry. However, it's important to note that no generative system is perfect on the first try. The brick façade, with its intricate layers of curvature and repetition, presented a particularly complex challenge.
Here's a point that might spark some debate: One of the most significant takeaways from 'Bendo Bricks' is the identification of the delicate threshold between computational precision and visual interpretation. Undale points out that generative tools still grapple with the nuances of minute detailing and dense surface patterns. If the brick units are modeled too large, the AI can interpret them easily, but the result can feel architecturally unconvincing. Conversely, if they are too small, the tool struggles to render them accurately. The true success of this workflow hinges on finding that sweet spot – where the geometry remains believable, the details are legible, and the final image aligns harmoniously with architectural intent. This balance is where the real potential of this approach lies.
Ultimately, the 'Bendo Bricks' project showcases generative AI image editing through a focused, single-image workflow in ComfyUI. The AI analyzes an uploaded image to generate a design set that stays incredibly close to the original input, offering a powerful new avenue for architectural exploration.
A Bit About the Visionary: Kedar Undale
Kedar Undale is more than just an architect; he's a parametric designer and a leading figure in computational art. Based at Kedar Undale Design Studio, he’s dedicated to consulting, education, and pushing the boundaries of computational art. His academic journey includes a master’s degree from the prestigious IAAC in Barcelona. Currently, he’s a ComfyUI Specialist for Madespace, a California-based company, and previously served as a Computational Designer at Partisans in Toronto. For the past eight years, Undale has been a global educator, leading international workshops on Parametric Thinking at renowned institutions like DigitalFUTURES and CEPT India. With the rise of generative design, he’s been at the forefront, developing custom workflows and training sessions for organizations such as PAACADEMY and FUTURLY. His artistic endeavors, featured in Homegrown Magazine India, explore the visualization of intangible forces like sound, wind, and gravity using computational systems.
Project Snapshot:
- Project Name: Bendo Bricks
- Lead Designer: Kedar Undale
- Key Tools: Rhino3D, Grasshopper3D, ComfyUI, Gemini 3 Pro Image (Nano Banana Pro), Adobe Premiere Pro
What do you think about this fusion of AI and parametric design? Does it excite you about the future of architecture, or do you have concerns about AI's role in creative processes? Share your thoughts in the comments below!
