AI, 3D and virtual sampling: fewer samples, faster decisions

KEY TAKEAWAYS Summary by the editors

1. Virtual sampling means reviewing a garment as a 3D simulation built from real patterns and fabric data, instead of producing a physical sample for every design iteration.
2. HUGO BOSS reported that around 75% of its products were developed digitally in fiscal year 2025, and says digital development further reduces the need for physical samples.
3. WRAP's Clothing Knowledge Hub cites adidas cutting product development time by a third, from 18 to 12 months, by replacing physical prototypes and sales samples with virtual 3D files.
4. AI adds speed around 3D, for example generating concepts, turning sketches into visuals and, in research, converting design inputs into sewing patterns, but accurate simulation still depends on measured fabric properties and correct patterns.
5. Physical samples remain necessary for new fabrics, complex constructions, handle and final fit approval, so the realistic goal is fewer sampling rounds, not zero samples.

Virtual sampling is the practice of reviewing a garment as a realistic 3D simulation, built from its actual pattern pieces and measured fabric properties, rather than producing a physical sample for each iteration. AI adds speed around that process, from generating concepts to automating parts of pattern work and rendering. Together they let brands make design, fit and range decisions with fewer physical samples, though not without any.

What is virtual sampling in fashion?

A traditional development cycle can involve several rounds of samples: a proto to check the idea, fit samples to correct the pattern, and salesman samples to show buyers. Each round means fabric, sewing time, shipping between supplier and brand, and waiting. In virtual sampling, the pattern is loaded into 3D software, fabric parameters such as weight, stretch and bending are applied, and the garment is draped on a digital avatar. Designers, technologists and merchandisers review the simulation, comment and request changes, and only proceed to a physical sample once the digital version is approved.

This approach predates generative AI. Tommy Hilfiger, for example, said in 2019 that it had been on its 3D design journey since 2017 and was targeting full 3D design for its apparel collections from spring 2022, according to FashionNetwork.

How does AI add to 3D and virtual sampling?

AI works at both ends of the 3D process. Upstream, generative tools turn sketches, mood boards and text into visuals; McKinsey's 2023 analysis of generative AI in fashion described designers converting sketches and mood boards into high-fidelity designs and generating variations from existing lines. Downstream, AI can speed up rendering, create on-model imagery and help produce variants in different colours and materials.

The step in between, turning an idea into a correct sewing pattern, is the hardest and still largely human. Research is advancing: Design2GarmentCode, presented at the CVPR 2025 computer vision conference, uses large multimodal models to convert photos, sketches or text into parametric pattern-making programs that produce sewing patterns. Such work is promising, but it is research, not yet a replacement for experienced pattern makers in commercial development.

Read also

AI in fashion product development: from brief to sample

How many physical samples can brands replace?

Published figures are scarce and usually brand-reported. HUGO BOSS states that around 75% of its products were developed digitally in fiscal year 2025, up from around 65% in 2024, using in-house image generators, virtual try-ons with avatars and immersive 3D simulations; it says this further reduces the need for physical samples without compromising performance and fit. WRAP's Clothing Knowledge Hub cites adidas replacing physical prototypes and sales samples with virtual 3D files, reducing product development time by a third, from 18 to 12 months. Tommy Hilfiger said in 2019 that men's dress shirts for Fall 2020 would be fully 3D designed and require no sample production.

These examples share a pattern: the biggest gains come in categories with stable blocks and well-understood fabrics, and from removing intermediate rounds rather than the final fit sample.

What does a virtual sample need to be trusted?

Of these inputs, fabric data is usually the bottleneck. A 3D tool needs measured values for properties such as stretch, bending and weight, which are typically captured with dedicated test equipment or provided by mills in a compatible format. Brands that ask core suppliers to deliver digital fabric files together with swatches build their library much faster than those that try to digitise everything in-house. Avatars are the second constraint: if the digital fit model does not match the physical fit model the brand trusts, technical teams will reasonably refuse to approve fit on screen.

What are the limits of virtual samples?

• Handle and hand-feel: no screen conveys how a fabric feels, which matters in premium and knitwear categories.
• New materials and constructions: without measured parameters and proven blocks, simulations are guesses.
• Fit on real bodies: avatars approximate; final fit approval is still normally done physically.
• Up-front effort: digitising fabrics, building avatars and training teams takes time and budget before savings appear.
• Supplier readiness: benefits shrink if suppliers cannot work from digital files.

How do virtual samples change buying and selling?

Once a style exists as an approved digital asset, it can be reused beyond development. HUGO BOSS says its digitally developed styles are used, for example, with wholesale partners via digital showrooms and for virtual try-ons for customers on its website. That reuse is where much of the return on investment sits: the same 3D asset supports range reviews, sell-in presentations, e-commerce imagery and, increasingly, AI-generated content in multiple colourways.

Read also

How is AI used for colour, prints and materials in fashion?

How should a brand get started with virtual sampling?

1. Choose one category with stable blocks, such as shirts, polos or jersey.
2. Digitise core fabrics with measured parameters and build avatars from your own fit models.
3. Run digital and physical samples in parallel for one season and compare decisions.
4. Track sample counts, shipments and calendar days per approved style.
5. Extend to more categories and to sales presentations once the data shows the simulation is trustworthy.

The business case should be judged over several seasons. Early savings in samples and shipping are often offset by the cost of digitising fabrics, building avatars and training teams. The return grows as the library of digital materials and approved blocks expands and as the same 3D assets are reused in sales and marketing.

Frequently asked questions

What is a virtual sample in fashion?

A virtual sample is a 3D simulation of a garment built from its real pattern and measured fabric data, draped on a digital avatar. Teams review and approve it on screen before, or instead of, producing a physical sample.

Can 3D virtual sampling replace physical samples completely?

Not completely in most categories. Brands report reducing the number of physical samples, especially intermediate rounds and sales samples, but fabric handle, new constructions and final fit usually still need physical checks.

How does AI improve 3D fashion design?

AI speeds up concept generation, turns sketches into visuals, accelerates rendering and creates variants. Research also explores generating sewing patterns from images or text, but accurate simulation still depends on correct patterns and measured fabric properties.

What does a brand need to start virtual sampling?

It needs digital patterns, measured fabric data, avatars matched to its fit models, trained designers and technologists, and suppliers able to work with digital files. A pilot in one stable category is the usual starting point.

GuideThe complete guide to AI in fashion design and product developmentRead the complete guide