AI + Industrial Design
Visualising vision with Vizcom
How CAD, mood references and a low-fidelity render helped an AI tool understand the product I had in mind.
For a Render Weekly birdhouse prompt, I wanted to see how far I could take a product visualisation using Vizcom together with familiar design tools.
The final images look simple, but the useful part was discovering what kind of input the tool needed before it could preserve the idea instead of reinventing it.
Find a direction
Move from broad references towards a product that can be reproduced.
Resolve the idea
Define the experience, construction and visual character of the birdhouse.
Build the input
Add form, materials, subject and environment to a low-fidelity reference.
Generate consistently
Use the same visual foundation to explore lighting and camera angles.
Finding the idea
I began by searching broadly—scrolling through Pinterest, assembling moodboards and testing several directions.
- A birdhouse inspired by the Gujarati chabutro
- A terracotta form using patterns from kolam art
- A three-quarter cylindrical birdhouse for a corner wall
- A lightweight birdhouse made from straw
Most of these felt closer to handcrafted objects. I wanted a design that could be manufactured repeatedly while still feeling warm and playful.
A transparent birdhouse offered the balance I was looking for: a simple manufactured object with a new experience inside it.
Intention behind the design
The transparent enclosure began with a question: what if people could observe birds even while they rested inside their nest?
Existing transparent birdhouses were rare, and the examples I found felt sterile. I chose acrylic for a clean structure, then introduced a familiar house silhouette, sheet-metal mounts and a wooden perch to give the object strength and character.
See inside
Transparent walls create a gentle view into the bird’s resting space.
Stay elevated
A suspended structure keeps the birdhouse above the ground.
Protect the nest
The product forms a roof and enclosure around natural nesting material.
I wanted the object to sit around a natural nest, rather than behave like an artificial replacement for one.
Three visualisation strategies
I had a clear atmosphere in mind: rain, soft natural light, transparent acrylic and a small bird inhabiting the object. The challenge was finding an input that communicated all of those elements without losing the product itself.
CAD model + text prompt
I made a quick model in Rhino and uploaded it with a basic prompt. I also tried sketching over it, changing the background and adjusting the wording.
The form was present, but the result was far from the intended scene.
CAD model + custom image palette
Next, I trained custom image palettes in Vizcom using atmosphere references generated in Ideogram. The mood improved, but the tool still struggled to interpret the bird, nest, transparent surfaces and metal parts consistently.
The palette shaped the atmosphere, but it could not resolve missing information.
Low-fidelity KeyShot render + Vizcom
Finally, I made a rough KeyShot render containing every important element: the product, bird, nest, transparency, background and approximate composition. It was deliberately unfinished, but it described the complete scene.
Once the reference carried the full intent, Vizcom produced the lighting, reflections and material character I wanted.
Consistent final visuals
I repeated the same process with different camera angles. Because the core form, subject and environment were already present in each reference, Vizcom could concentrate on visual quality instead of guessing the scene.
Review
A rough render can be a precise instruction
The final workflow used each tool for a different kind of control. CAD established the product, KeyShot assembled the scene, Ideogram helped define the atmosphere and Vizcom developed the final visual character.
The low-fidelity render was the turning point. It did not need to look finished; it needed to make the relationships between the form, materials, subject and environment unambiguous.
AI visualisation became useful when the input expressed the complete design intent.
