3D Tour Build Guide
iPhone To Browser Link, 30 Minutes
This guide walks through capturing a photorealistic 3D tour of any space with an iPhone, processing it into a 3D model, and publishing it as a shareable browser link. About 30 minutes of active work, hours of automated processing. Real estate agents are already using this to replace $2,000 photography sessions. The deeper point: AI is compressing the cost of services that used to be expensive across dozens of categories. The opening isn't "industries are getting eaten." The opening is for the people who learn the new workflow first and use it to deliver the same quality at a fraction of the price. That's the real story.
A real estate agent in my network just stopped paying for listing photography.
She used to spend about $2,000 per listing on a professional photographer plus a 3D Matterport scan. Two-week turnaround, half-day shoot, the works. Last month she scanned three properties herself with an iPhone, processed them through one of the new AI-powered tools, and had photorealistic browser-based 3D tours live the same evening. Total cost: under $50 in software credits. Total time: about 30 minutes of active work per listing.
The work isn't worse. In some ways it's better. Clients can scroll through the space on their phone, look around freely, walk room to room. The kind of experience that used to require a $40,000 Matterport rig is now a free app and a phone.
That single example is what every industry is going to look like in the next two years. AI is collapsing the cost structure of services that used to need specialized labor and expensive hardware. Photography. Video editing. Translation. Legal research. Basic accounting. Customer support. Music production. Anyone offering these services at the old price is going to lose to someone offering the same quality at a fraction of the price.
The framing matters. This isn't "AI eats real estate photography." That's lazy. The actual framing: the people who leverage AI first, and use it to deliver the same outcome for way less, are going to take the work from the people who don't.
This guide is the proof of concept. A 3D tour, captured on a phone, hosted on a browser link, in 30 minutes. If you understand what's happening here, you understand the pattern. The pattern is the actual lesson.
What you'll need
- An iPhone (12 or newer recommended for the LiDAR sensor, though some apps work on older phones)
- The three tools below
- A space to capture (your apartment, an office, a property, anything with walls)
That's it. No DSLR, no tripod, no lighting kit, no Matterport rig.
The tool stack
Three tools, all with free tiers. The whole pipeline costs $0 to start.
- Polycam (iPhone capture). Free tier works for unlimited captures. Paid tier ($15/month) gives you more exports and higher resolution. Use this to capture the space.
- SuperSplat (browser-based cleanup). Free, no signup required. This is where you clean up your raw scan and compress the file. Open at superspl.at/editor.
- PlayCanvas (hosting and embedding). Free tier covers public projects, $15/month for private. Only needed if you want to embed the tour on your own website. Otherwise SuperSplat publishes a shareable link for you.
Other options exist (Luma AI, RealityScan, KIRI Engine), but this is the cleanest workflow if you want a polished, embeddable result. The principles transfer if you prefer a different tool.
The 5 capture rules
Most failed 3D tours fail because of the capture, not the software. These five rules are what separate a usable scan from a mess.
The rest of this playbook is below.
Enter your email to read it. You also get one new playbook every week.
One email a week. Unsubscribe any time.
- Overlap your sweeps. When you change rooms or angles, overlap your previous sweep by about 30%. The app needs the visual continuity to stitch the rooms together correctly. Skip the overlap, get a broken model.
- Stay at chest height, mostly. Keep the phone roughly at chest level for the main sweep. Hold it steady, slowly pan. Then do a second pass with the phone slightly higher and a third with it slightly lower if you want full vertical coverage.
- Capture in even lighting. Bright sunlight pouring through one window while another corner is dark will confuse the model. Either close blinds and use interior lighting, or pick a time of day when lighting is even.
- Don't capture moving things. Pets, kids, people walking through, ceiling fans, anything that moves. The model averages frames and moving subjects become smears. Empty the space or pause the capture while they move.
Follow these and your first capture will look like your fifth. Skip them and you'll spend hours on re-shoots.
The 4-step workflow
Step 1: Capture in Polycam
Open Polycam on your iPhone, hit "Start a scan," walk the space following the 5 rules above. A standard apartment takes 5 to 8 minutes. A house takes 10 to 15.
When done, hit "Process." Polycam uploads to its servers and runs the photogrammetry. Takes 10 to 15 minutes on a typical scan. You don't need to be at your phone during this part.
Step 2: Export the splat
Once processing is done, export your scan as a Gaussian splat (.ply file). Polycam offers this on the free tier. Download it to your computer. The raw file is usually 100 to 250MB depending on the size of the space.
Step 3: Clean the splat in SuperSplat
Raw splats have garbage. Floaters in the air, splats outside the house, blurry edges. SuperSplat fixes this in 5 minutes.
Open superspl.at/editor. Drag your .ply file in. Three tools you'll use:
- Bounding box crop. Drag the box around just the house. Delete everything outside.
- Lasso tool. Select floaters in the air (the wispy clouds above the ceiling). Press delete.
- Compression export. File → Export → Compressed .ply. A 200MB raw splat drops to 35MB with no visible quality loss.
Save the cleaned, compressed file. This is what you'll publish.
Step 4: Publish and share
SuperSplat has a built-in "Publish" option that gives you a hosted shareable link. Click Publish → get a URL → send it to anyone. They open it in any browser. No app, no install, no VR headset. They walk through with finger or mouse.
If you want more control, upload your .ply to a PlayCanvas project (free tier covers public projects, $15/month for private). Then embed it in any website with a one-line iframe.
Total cost so far: $0 (Polycam free tier + SuperSplat free + PlayCanvas free).
The business layer
This is what nobody talks about in the standard guides for these tools.
If you understand this workflow and traditional service providers in your area don't, you have a real opening. A few categories where the gap is obvious:
- Real estate agents still paying $1,500 to $3,000 per listing for photo + 3D tour packages. Many are running on a workflow that hasn't been updated in five years.
- Airbnb and short-term rental hosts who'd love a 3D tour of their property but can't justify the photography cost.
- Small businesses (restaurants, gyms, retail, salons) that would benefit from a 3D walkthrough on their website but never get one because of cost.
- Event venues that need to show couples and corporate clients what the space looks like without an in-person tour.
- Property managers who manage multiple units and could differentiate on listing quality.
What this looks like in practice: someone in your network is paying $2,000 for a service. You can deliver the same outcome for $300 to $800 and still be making strong margin. They get a 60 to 80% cost reduction. You get a service business with virtually no overhead.
A few honest notes before you go.
This isn't a get-rich-quick play. It's a service business. You still need to acquire clients, deliver consistently, handle revisions, manage expectations. The 30-minute capture is the easy part. Running a service business is the actual work.
Quality matters more than ever. Because the barrier to entry just dropped, the only thing that differentiates you is the quality of your output. Spend time on the capture. Re-shoot if it's not right. Don't ship a model with gaps just because it was free to make.
In some regions, real estate photography is regulated and tours used in MLS listings have specific technical requirements. Check before you charge for MLS-bound work.
The trade itself isn't proprietary. Anyone can learn this in an afternoon. Your edge is delivery, consistency, and the relationships that drive repeat business. Same as any service business that ever existed.
The bigger pattern
The 3D tour is one example. The pattern is the actual story.
AI is compressing the cost structure of dozens of services that used to require specialized labor or expensive hardware. Photography. Video editing. Voiceover work. Translation. Bookkeeping. Legal research. Tutoring. Music production. Logo and brand design. Architectural visualization. Basic content writing.
Every one of those categories has the same dynamic playing out. Old workflow: expensive, slow, gated by tools and skills. New workflow: cheap, fast, accessible to anyone willing to learn the tool. The work isn't worse. In most cases it's faster and more flexible.
Two things are true about what happens next.
First, the people offering the old workflow at old prices are going to lose share to people offering the new workflow at lower prices. This isn't a prediction. It's already happening in every category I just listed. The question is how fast it happens in your specific niche.
Second, the opportunity isn't on the side of the established service providers. It's on the side of the people who learn the new workflow first and position themselves as the modern alternative. The first wave of these new service operators is going to take meaningful share before the incumbents adapt.
If you're already running a business, the question is: which AI workflow could you adopt that lets you deliver your current service for half the cost, or your competitor's service for a fraction of theirs?
If you're not running a business yet and you've been looking for an opening: every one of these compressed-cost categories is one. Pick the one that matches your interests, learn the workflow deeply, find your first three clients, build from there.
The asymmetry isn't between humans and AI. It's between people who learn the new workflows and people who don't.
