How catalog operators should think about 3D furniture modelling in 2026, and when AI 3D generation from a single product photo replaces traditional CGI.
3D models have quietly become a default catalog asset for furniture brands — a baseline expectation for AR, configurators, marketplace 3D viewers, and lifestyle scene generation. The economics, however, have changed. AI 3D generation from a single product photo now produces GLB/glTF/OBJ output in minutes, where traditional 3D modelling studios still quote in days or weeks. This guide is for catalog operators who need to make a sober, vendor-neutral decision about how to model their range — and how platforms like Furniture Connect fit into that decision.
Furniture is the category where 3D pays back fastest. The buyer is committing to a large, infrequent, physical purchase; conversion lifts from interactive 3D and AR are well-documented, and Google's commerce guidance now treats 3D assets as first-class product media — per Google Search Central's merchant guidance, listings with embedded 3D are surfaced differently than flat imagery.
Four pressures are driving 3D adoption at catalog scale:
model/gltf-binary (GLB) directly in product viewers. Brands without 3D show up as the flat-image option in a 3D-first carousel.Brands that resisted 3D in 2022 cited cost. That objection no longer holds in 2026 — the cost curve has flipped.
A 3D model is not a deliverable — it is an input to many downstream assets. Understanding the downstream uses is what tells you which fidelity, format, and topology you actually need. Most operators over-spec, paying studio prices for cinema-grade meshes when channels truncate textures to 2K anyway.
The realistic uses for a catalog 3D asset:
Each use cares about different parts of the asset. A perfect AR model can be a useless print model, and vice versa. The choice of provider is downstream of the choice of use case.
Traditional 3D modelling — produced in-house with desktop CAD applications, outsourced to a CGI studio, or sourced from a 3D marketplace — has been the default for fifteen years. It still produces beautiful work. It also still costs what it cost in 2018.
Industry rate cards from studios serving the furniture sector typically fall in this range:
| Asset type | Typical price per SKU | Typical turnaround |
|---|---|---|
| Photoreal hero model (cinema-quality) | £400–£1,200 | 2–4 weeks |
| Configurator-ready PBR model | £250–£600 | 1–3 weeks |
| AR-optimised GLB/USDZ pair | £150–£400 | 1–2 weeks |
| 3D marketplace pre-built model (close match) | £20–£150 | Hours |
| Revisions and consistency rounds | 15–30% surcharge | +1 week |
These are vendor-neutral medians from public studio price sheets and freelancer marketplaces — the bill operations actually pays after revisions.
Where the traditional model breaks:
None of these are flaws of any particular studio. They are structural properties of the human-labour 3D production model — what AI 3D generation was built to disrupt.
AI 3D generation is the technique of producing a textured 3D mesh from one or more product photos, using a mix of underlying AI models with intelligent routing. The output is a standard 3D asset — GLB, glTF, or OBJ — that drops into the same downstream pipelines a hand-built model would feed.
Furniture Connect's AI 3D generation works from a single product photo and returns GLB or glTF in minutes rather than days or weeks. The platform is purpose-built for furniture catalogs, pairing the 3D output with PIM write-back, DAM storage, and downstream lifestyle scene generation in one workflow.
The economic shift is genuine. A SKU that cost £300 and two weeks now costs pennies and minutes. That order-of-magnitude change is what unlocks catalog-wide 3D coverage instead of hero-SKU coverage.
What AI 3D generation does well today:
What it does less well:
The honest framing is: AI 3D generation is now the right default for the bulk of a furniture catalog, with traditional 3D reserved for the cases that genuinely need it.
Format confusion is the single biggest reason catalog 3D projects stall. In reality there are only four files an operator needs to understand, and three are governed by the same open standard — the Khronos Group's glTF specification, the dominant open 3D format for the web.
.gltf + textures). The "JSON for 3D" format. Human-readable, slightly larger on disk, useful for editing and inspection. Most pipelines convert from glTF to GLB before shipping..glb). The binary, single-file packaging of glTF. This is the format almost every modern web viewer, AR system, and marketplace viewer expects. If you only buy one format, buy GLB..usdz). Apple's preferred AR format, derived from Pixar's USD. Required for iOS Quick Look AR. Most AI 3D generators output GLB and convert to USDZ as a downstream step..obj + .mtl). The old industry workhorse. Useful for compatibility with legacy 3D pipelines and for editing in traditional 3D tools. Larger files, no animation, no PBR materials natively. Still asked for by some print and CAD pipelines.Other formats — FBX, STL, STEP — are CAD-pipeline assets, not catalog assets. If a buyer requires them, they belong to a different workflow.
Rule of thumb: ship GLB for web and Android AR, ship USDZ for iOS AR, keep glTF or OBJ on hand for downstream editing, and stop worrying about everything else.
This guide is not anti-traditional-3D. Three categories still favour traditional 3D modelling, and operators who ignore the nuance ship worse catalogs.
For everything else — the eighty to ninety percent of a catalog that is in-production SKUs with available photography — AI 3D generation is now the default. We have argued the same logic from the imagery side in the anatomy of a perfect product listing.
Operators do not buy 3D models; they buy catalog coverage at a target cost per SKU. Re-framing the decision this way changes which questions matter.
The honest unit-economics comparison:
| Approach | Cost per modelled SKU | Time per SKU | Coverage at £100k budget |
|---|---|---|---|
| Traditional CGI studio (hero quality) | £400–£1,200 | 2–4 weeks | 80–250 SKUs |
| Traditional CGI studio (configurator quality) | £250–£600 | 1–3 weeks | 165–400 SKUs |
| In-house team using traditional 3D tools | £150–£400 (fully loaded) | days | 250–650 SKUs |
| 3D marketplace pre-built models (close-match only) | £20–£150 | hours | 650–5,000 SKUs (if matches exist) |
| AI 3D generation from product photo | Pennies to a few pounds per SKU | minutes | Whole catalog |
The McKinsey retail operations research on furniture and home goods has repeatedly identified content production as the single largest line item in ecommerce launch cost. Baymard's usability research on product-page design also notes that interactive 3D and AR availability materially affects perceived product confidence — which is what catalog 3D is ultimately buying.
When AI 3D generation collapses the per-SKU cost by two orders of magnitude, the operator's question stops being "which SKUs deserve 3D" and starts being "what additional uses can we now afford." That is the economics shift. You can run the numbers against your own catalog before committing, and check pricing against your volume.
A pilot is not a proof of concept. It is a structured comparison against a known baseline. Three weeks is enough to learn whether AI 3D works for your catalog.
Week 1 — baseline.
Week 2 — generation and review.
Week 3 — channel test.
The pilot's output is a coverage policy: which categories run AI 3D by default, which run traditional, which run hybrid. That policy is the deliverable. The 3D models are the byproduct.
For brands that have already run this pilot, the result is consistent across our case studies: 80%+ of catalog 3D moves to AI generation, hero SKUs and mechanical products stay on traditional pipelines, and total catalog cost falls by roughly an order of magnitude.
The structural point is simple. Traditional 3D modelling was built for an era when human-labour CGI was the only way to produce a usable mesh. That era is over for catalog-volume work. AI 3D generation does not replace traditional 3D — it absorbs the routine ninety percent so traditional 3D can be reserved for the ten percent that genuinely warrants it.
The operators who win the next three years will be the ones with the best coverage policy, clear about which SKUs need which approach. The fastest path to a working policy is to book a demo and walk through your own SKUs with our team.
For wider context, see Furniture Today for industry coverage, Shopify Plus for the commerce-platform perspective, and Statista's furniture datasets for volume trends.
An operator's framework for evaluating AI image providers for furniture catalogues — covering fidelity, workflow fit, PIM integration, unit economics, and rollout.
A technical due-diligence checklist for ecommerce operators evaluating AI image platforms for furniture catalogues: integration, governance, unit economics, and a 14-day pilot plan.
Operator's guide to AI furniture imagery for B2B showroom catalogs, lookbooks, sales-app screens, and dealer portals — workflow, governance, and rollout.
Join hundreds of furniture brands already using FurnitureConnect to launch products faster.