Introduction
After having done a bunch of house hunting (see property-hunt post), my partner and I settled on a condo near our workplaces. After a several-week process of reaching out to property agents and going for viewings (can't digitize this yet, unfortunately!), we made an offer for a 1BR + study apartment in V on Shenton. The next mini-project became clear then: planning furnishings and layouts of our upcoming house!
There are several considerations to be made, given that this is a 2-year rental arrangement and not a 'forever home':
- No permanent fixtures are allowed
- No wall mounts, or smart switches, etc.; any 'smart home' functionality has to be noninvasive (e.g. IR emulation)
- It should be robo-vacuum friendly to save us lots of cleaning trouble
- Wiring is already done up so there are no networking or cabling decisions to be made (sadly)
- We'll often have to have both of us working at the same time; our setup needs to accommodate that
There are several problems with answering our questions:
- How do we know whether our (relatively small) study room can fit two working setups?
- How do we know how any proposed furniture will look before we buy it? Will it even fit in the space that we theorize?
- How will different 'arrangements' of furniture look? How can we visualize this without buying and physically rearranging everything?
Typically the answer would be to pay an interior designer some lavish sum and get artists to do mockups. But I'm an engineer with some amount of free time! So here we go :-)
Process
The 'right' approach to do this would probably be a combination of Blender and 3DS Max - honestly, if I was starting this from scratch with no skill background I'd do it this way. However, during my time in SUTD I did plenty of 3D modelling in SolidWorks - their education license stickiness strategy has worked! - so SolidWorks became my lowest-friction, lowest-barrier-to-entry way of solving this (estimating a few hours in SolidWorks vs. likely a few days of UX learning on other platforms).
It's pretty clear that SolidWorks isn't the best software for this; it's meant more for electro-mechanical engineering tasks than for interior design. However, it's enough for me! The main things I need:
- Ability to combine multiple components (SolidWorks assemblies)
- Ability to 'project' images (floor plans) and trace contours/edges
- Ability to design individual components based on dimensions I can farm from different sources (e.g. furniture spec sheets)
- In the far future - LLM support :-( to auto-generate components
- Basic visualization capabilities (Honestly SolidWorks Visualize is pretty disappointing so far....)
- We only really need to focus on spaces where there are significant design choices to be made; no real need for toilet or bedroom design given the limited space
- Ability to leverage existing community-made components (e.g. Herman Miller chair 3D models), while checking individual usage licenses
Community Parts
Happily, I was able to find many fitting 3D models already built online - for example:
This made it much easier to create correctly scaled (and good-looking!) representations in our model.
Apartment
We begin by taking the floor plan - in this case, it's conveniently available from the V on Shenton floor plan page (which listed 1-bedroom + study unit types when I checked on 2026-04-06). We also did a bunch of physical measurements during one of our viewings so we could scale the floor plan to the correct dimensions for a proper scale model.
Furnishings
Scenarios
Having built all our components, we can now dynamically rearrange parts in our assembly!
Arrangements
After creating necessary assets, we can arrange them in our apartment!
Mini side project
The prototype code is now public at viz-on-shenton.
In theory, nothing's really stopping one from simply uploading a floor plan and getting an LLM to handle visualizing layouts and interacting with the user throughout. In practice, this still wasn't trivial to build robustly in my tests: everything from furniture design (significant back and forth with an LLM to confirm dimensions, textures, shape, etc.), to lighting, to generating 3D models, to placing 3D models within 3D models needed much more glue code than expected.
Some initial experimentation (1-2 hours) looked promising, but showed that a loooooot more work was needed - likely on the order of weeks to months for solo-dev-me, even with LLM assistance. I hence decided to fall back to my tried-and-true method of modelling it myself - sad but the Technology Just Isn't There Yet.
Wrap-up
This project did exactly what I needed for this rental: fast "will this fit?" iterations before spending money on furniture, plus a clearer view of tradeoffs in the study and living room. If I revisit the automated route later, I'll likely keep this manual model as the reference baseline to evaluate generated outputs against.