Reading time: 9 minutes.
There are four production paths fashion brands use to create 3D assets for virtual try-on in 2026: Digital Product Creation (DPC) files made during design, photogrammetry capture of physical samples, manual 3D sculpting by an artist, and AI-driven 2D-to-3D pipelines that convert standard product photos into digital twins. Each was built for a different starting point. Most of the industry confusion — and most of the cost — comes from treating these four as interchangeable when they are not. This piece walks through what each pipeline actually is, what it costs, what it requires, and when it is the right choice for a fashion brand.
Fashion brands keep being told they need 3D modelling to do virtual try-on. They don't. They need digital twins.
Those two phrases have been treated as synonyms for a decade, and the conflation is the single biggest reason AR try-on has stayed expensive. A 3D model is a production method — a file built by a designer or artist using specific software. A digital twin is an outcome — a 3D representation of a real product that can be placed on a real shopper in augmented reality. You can produce a digital twin with a 3D model, but you can also produce one without ever opening a 3D modelling tool.
That distinction is what this article is about. Below are the four production paths in active use today, the realistic numbers behind each, and how a fashion brand should decide between them.
WEARFITS is an AI-powered virtual try-on platform for footwear, bags, and apparel. We built our own pipeline because we ran into the conflation problem first-hand — and the rest of this article explains why each of the other three pipelines exists, what they do well, and where each one breaks. You can see the WEARFITS 2D-to-3D approach in action on our photo-to-AR demo page.
What it is. Digital Product Creation, usually shortened to DPC, is the design-stage 3D workflow used inside fashion product development. Designers and pattern engineers use tools like CLO3D, Browzwear, and Style3D to build a virtual version of a garment or shoe while the product is being designed — before any physical sample is made.
DPC was not built to produce AR try-on assets. It was built to reduce the number of physical samples a brand has to make, accelerate time to market, and give designers a way to iterate on fit and construction digitally.
According to Browzwear's own explainer, DPC reduces physical sampling, accelerates time to market, and produces 3D design files that can be reused downstream — for marketing renders, e-commerce visuals, and, with extra work, for AR.
What it costs and how long it takes. DPC cost is sunk inside the product development budget. If you already run DPC for design, the marginal cost of having a 3D file to start with is effectively zero. If you don't run DPC, retrofitting it onto an existing catalogue is not feasible — DPC files are made at the point of design, not after the fact.
Where DPC breaks for AR try-on. Three places. First, DPC files are usually rigged for design simulation, not for mobile AR — they need re-meshing, texture optimisation, and rigging work before they can ship to a storefront. Second, DPC adoption inside fashion is still partial; most footwear and accessory brands do not yet have a DPC file for every SKU in their live catalogue. Third, even brands that do run DPC tend to use it on hero categories — apparel, technical garments — not on the full long tail of bags, smaller leather goods, or seasonal footwear drops.
When DPC is the right choice for virtual try-on. When a brand already runs DPC at scale, has design files for the SKUs it wants to AR-enable, and is willing to invest in a separate AR rigging pipeline on top.
What it is. Photogrammetry is a capture-based method. The brand puts a physical sample of the product on a turntable or static rig, photographs it from dozens of angles, and feeds those photos into reconstruction software (RealityCapture, Metashape, Meshroom) that triangulates the geometry from overlapping images.
The output is a high-resolution 3D mesh that is dimensionally accurate to the original physical object — assuming you measured carefully and lit it well.
How many photos and how long. OpenScan's benchmarking finds the optimal count for turntable photogrammetry is around 200 photos per object. Below 100 the mesh quality drops; above 300 the processing time grows without improving the result.
End-to-end, a single SKU usually takes one to three working days. Capture is a few hours; processing on a workstation is another few hours; mesh cleanup and texture baking is the slowest step — that's where the human time goes.
What it costs. Indicative cost per SKU is in the range of $150 to $500 once you factor in capture rig time, software, and a technician to do the mesh cleanup. Brands that bring photogrammetry in-house pay a CapEx cost upfront (rig, lighting, scanner, software) and a per-SKU operating cost in technician time.
Where photogrammetry breaks for virtual try-on at catalogue scale. Three places. First, you need the physical sample in your photogrammetry studio, which is a logistics constraint when products are made overseas or have not yet shipped. Second, the raw output is too heavy for mobile AR — most reconstructed meshes are millions of polygons and need optimisation before they can render in a browser. Third, the cost-per-SKU and time-per-SKU mean photogrammetry stays viable for tens of products, not thousands.
When photogrammetry is the right choice. When a brand needs scan-accurate digital replicas of a small set of premium hero products — typically high-value items where the dimensional fidelity is worth the per-SKU cost.
What it is. A senior 3D modeller — usually a freelance specialist or an in-house artist — opens Blender, ZBrush, or Maya, uses product photos as reference, and sculpts the product from scratch. The artist builds the mesh, paints or bakes textures, sets up materials, and rigs the result for whatever downstream use is needed.
This is the oldest pipeline of the four. It produces high-quality results in the hands of a good artist. It does not scale.
How long and what it costs. Sketchfab's enterprise breakdown puts simple 3D models at under $200 with around 2 hours of artist time, medium complexity at under $650 with around 15 hours, and high complexity at $650+ with 40 or more hours. For a typical footwear SKU rendered to a quality acceptable for AR try-on, the realistic range is 15 to 50 hours of artist time per SKU and $200 to $800 per SKU at typical freelance rates.
Where manual modelling breaks for virtual try-on at catalogue scale. The math. At $400 per SKU and three days of artist turnaround, a 500-SKU catalogue is a $200,000 six-month project before the first try-on ships. For most fashion brands that is not a 3D modelling project; it is a small-equity investment.
When manual modelling is the right choice. When a brand wants a small number of hero SKUs — a handful of icon products, a flagship campaign drop — produced with bespoke art direction that an AI pipeline will not match. The output is custom by definition, and the cost reflects that.
What it is. An AI-driven 2D-to-3D pipeline takes standard product photos — the same packshots already used on product detail pages — and uses generative AI to produce a 3D digital twin of the product. No physical sample is required. No 3D artist is required. No CAD or DPC file is required.
WEARFITS, a web-first virtual try-on platform founded in Krakow, built its 2D-to-3D pipeline specifically to address the scaling problem the other three pipelines run into at catalogue size. The input is four standard packshot photos of a product against a clean background. The output is an AR-optimised digital twin, ready for mobile try-on, with realistic occlusion and fit calculation. You can see the workflow on our photo-to-AR demo page.
How long and what it costs. Asset generation runs in minutes per SKU. A 100-SKU catalogue runs end-to-end — from photo upload through to live AR try-on on the storefront — in under two hours in production environments. Cost is built into a subscription model rather than priced per SKU, which inverts the economics of the other three pipelines.
Where the AI 2D-to-3D pipeline breaks. Honest answer: the AI pipeline is best for products with reasonably standard form factors (footwear, bags, structured accessories, defined apparel silhouettes) where the brand has clean packshot photography. It is less suited to highly bespoke or unusual product geometries, products with extreme transparency or reflective surfaces, or products where pixel-perfect dimensional fidelity matters more than visual realism — in those cases photogrammetry or manual modelling still wins.
When the AI 2D-to-3D pipeline is the right choice. When a brand needs realistic AR try-on across an entire catalogue, fast, using the photography it already has. WEARFITS deploys across Shopify, WooCommerce, mobile web, and in-store AR mirrors from a single integration — one production pipeline, every channel.
| DPC | Photogrammetry | Manual 3D modelling | AI 2D-to-3D (WEARFITS) | |
|---|---|---|---|---|
| Source input | Design-stage 3D files | Physical sample + ~200 photos | Product photos + 3D artist | 4 standard packshot photos |
| Skills required | DPC software, pattern engineer | Photogrammetry software, capture rig, mesh cleanup | Senior 3D modeller and texture artist | None |
| Time per SKU | Sunk in design phase | 1–3 working days | 15–50+ artist hours | Minutes |
| Time for 100 SKUs | Only if all 100 already exist | Weeks to months | Months | Under 2 hours |
| Indicative cost per SKU | Sunk in design budget | $150–$500 | $200–$800 | Subscription-based |
| AR-ready out of the box | No, needs rigging | No, needs optimisation | Varies | Yes |
| Best when | You already run DPC at scale | Small set of scan-accurate hero SKUs | Bespoke art direction on a few SKUs | Realistic AR across the entire catalogue, fast |
The decision tree is shorter than the comparison table suggests:
The honest answer for most fashion brands is "a combination." A few hero SKUs justify photogrammetry or manual modelling. The other 95% of the catalogue does not. WEARFITS deploys across Shopify, mobile WebViews, and in-store AR mirrors from one integration, which means a brand can run a hybrid stack — manual for hero, AI for tail — without having to manage four production pipelines in parallel.
The economics of virtual try-on changed in 2025 and 2026, and the production pipeline is the reason.
For most of the past decade, AR try-on lived in a particular budget category — six-figure project, six-week turnaround, six SKUs. That was a function of the production pipeline being either DPC-dependent, photogrammetry-dependent, or artist-dependent. Each of those has good reasons to exist. None of them was built to scale across thousands of SKUs at fashion's actual catalogue depth.
The arrival of AI 2D-to-3D pipelines didn't make the other three obsolete. It separated the production method from the outcome. The outcome — a digital twin of a product, usable in mobile AR, optimised for browser performance — can now be produced from photos a brand already has. That changes the budget category virtual try-on lives in.
WEARFITS is an AI-powered virtual try-on platform that deploys across Shopify, WooCommerce, mobile WebViews, and in-store AR mirrors from a single integration. The 2D-to-3D pipeline is the production method underneath it; the digital twin is the asset; the AR try-on is the outcome a shopper sees on a phone. Keeping the three concepts separate is what makes the rest of the decision-making clear. The clearest way to see what that means in practice is the photo-to-AR demo.
A 3D model is a production method — a file built in modelling software. A digital twin is an outcome — a 3D representation of a real product that can be rendered in AR. You can produce a digital twin using a 3D model, but you can also produce one using photogrammetry, AI 2D-to-3D, or DPC files. The two terms have been used interchangeably for a decade; keeping them separate is what makes pipeline choice easier.
Cost depends entirely on the pipeline. Photogrammetry runs $150–$500 per SKU including capture and cleanup. Manual 3D modelling by an artist runs $200–$800 per SKU at typical freelance rates. DPC has no marginal cost if you already run it for design, but full retrofit is not viable. AI 2D-to-3D pipelines like WEARFITS price on a subscription rather than per SKU.
An AI 2D-to-3D pipeline. WEARFITS deploys up to 100 SKUs in under two hours in production environments, using standard packshot photography as input. The other three pipelines — DPC, photogrammetry, and manual modelling — are not built to produce catalogue-scale output in those timeframes. The photo-to-AR demo shows the workflow end-to-end.
For photogrammetry, yes — it is a capture method that requires the physical product on a rig. For manual 3D modelling and for AI 2D-to-3D, no — both work from product photos. For DPC, the question doesn't apply because DPC files are made during product design, before physical samples exist.
Photogrammetry is the process of taking many photos around a physical object — usually 100 to 200, sometimes more — and using reconstruction software to triangulate a 3D mesh from the overlap between the images. It produces dimensionally accurate models, but it requires the physical sample, capture hardware, and cleanup time.
DPC, or Digital Product Creation, is the 3D design workflow fashion brands use during product development using tools like CLO3D, Browzwear, and Style3D. It produces 3D garment files at the design stage. Those files can be repurposed for AR try-on with additional rigging work, but DPC is not itself a virtual try-on pipeline — it is a design pipeline that happens to produce assets that can be reused downstream.
Not for every use case. AI pipelines are strongest for products with reasonably standard form factors and clean packshot photography — footwear, bags, structured accessories. For unusual geometries, transparent or highly reflective materials, or hero pieces that need bespoke art direction, manual 3D modelling still wins. The right approach for most fashion brands is a hybrid: AI for the catalogue, manual for hero pieces.
Try the WEARFITS photo-to-AR workflow on the demo page, or install the WEARFITS Shopify app to deploy AR try-on across your catalogue.