Virtual sampling is the use of digital garment samples, made with 3D software or AI image generation, in place of some physical samples during product development. Factories use it to win design approval, choose fabrics and colourways, and settle early fit questions before cutting cloth. Physical samples remain for final fit, pre-production and shipment approval.
Sampling is one of the slowest and most wasteful parts of the garment business. Every round means cutting fabric, stitching, couriering the sample to the buyer and waiting for comments. Many samples are rejected on looks alone, which is exactly the part a digital sample can show. This guide sets out where the digital tools fit, where they do not, and what the published evidence actually says.
What are the stages of physical sampling?
| Sample | Purpose | Typical question it answers |
|---|---|---|
| Proto (development) sample | First interpretation of the design | Does the style look like the sketch? |
| Fit sample (often several rounds) | Check fit on a fit model or form | Do the measurements and proportions work? |
| Salesman or showroom sample (SMS) | Show the range to retail buyers before orders | Will buyers order this, and in which colours? |
| Lab dip and print strike-off | Approve dyed colour and printed artwork on the real cloth | Is the colour and print right on this fabric? |
| Size set | Check grading across the size range | Does every size fit to spec? |
| Pre-production (PP) sample | Final approval with production fabric and trims | Is this exactly what will be made in bulk? |
| Top of production (TOP) or shipment sample | Taken from the bulk run for quality checks | Does the bulk match the approved PP sample? |
Names and the number of rounds vary between buyers and factories; this is the common sequence.
What is the difference between 3D and AI virtual sampling?
3D virtual sampling uses software such as CLO 3D, Browzwear or Style3D. A pattern maker builds the garment from real digital patterns, assigns measured fabric properties (stretch, bending, weight) and drapes it on an avatar. The output is technically meaningful: you can see where a garment pulls or gapes. The cost is the setup: licences, trained 3D staff, digitized patterns and fabric testing.
AI image sampling starts from a photo of the fabric and a description of the garment. An image model renders a photoreal garment in that cloth, usually within a minute. It needs no patterns and no specialist staff, so a merchandiser or a fabric salesperson can use it. It shows how the fabric looks as a style, not how the pattern fits a body: drape and print scale are interpreted rather than simulated.
The two are complementary. AI imagery is strongest at the very front of development, where the question is "which of these 40 fabrics should become which styles?" 3D is strongest in the middle, where the question is "does this pattern fit?"
Which sampling stages can virtual sampling replace?
| Stage | AI image preview | 3D simulation | Still needs physical? |
|---|---|---|---|
| Concept and design approval | Yes, well suited | Yes | Rarely |
| Fabric and colourway selection | Yes, well suited, many options fast | Yes | Colour sign-off still needs a lab dip |
| Proto sample | Partly: shows the look, not the pattern | Yes, from real patterns | Often one physical proto for hand feel |
| Fit rounds | No | Partly: early rounds | Yes, final fit on a body |
| Salesman sample and line review | Partly: images for decks and pre-selling | Partly: renders for decks | Buyers who must touch the fabric |
| Lab dip and strike-off | No | No | Yes |
| Size set, PP and TOP samples | No | No | Yes |
How much time and cost does virtual sampling save?
Published figures are few, and most come from software vendors, so read them as indications from specific programmes rather than industry averages.
The Australian Fashion Council's FashTech Lab ran its first digital sampling pilot in 2022. It reports that producing 30 digital samples in place of physical ones delivered a 50% reduction in sampling costs, a fall in sampling time from 12 weeks to 4, and up to 450 metres less textile waste.
Australian group APG & Co, owner of Saba, Sportscraft and Jag, reported a 50% reduction in development and fit sampling costs and an 80% time reduction on print strike-offs in its 2025 financial year, after several years of 3D adoption. The figures are reported by CLO Virtual Fashion, the maker of the software APG & Co uses, citing Ragtrader.
Both are 3D programmes run by established brands. There is no comparable published study yet for AI image sampling, so we do not quote one. The mechanism is simpler to reason about: every proto or salesman sample that a digital image makes unnecessary saves its fabric, stitching and courier time, and every rejected idea caught on screen never reaches the cutting table.
How to start virtual sampling in a factory
- 1
Pick one bottleneck, not the whole process
The easiest win is usually the front end: showing a buyer what new fabrics look like as garments before any sample is stitched. Leave fit and production approval as they are.
- 2
Digitize your fabric library
Photograph each fabric flat, in daylight, with a full print repeat in frame. The fabric photo guide explains the method. In Preloom, saved swatches carry their name and composition and cost nothing to store.
- 3
Generate the range as a grid
Instead of one style at a time, generate fabrics against styles together. A Preloom batch grid takes up to 12 fabrics and up to 12 garment styles and produces every combination, up to 60 designs, in the background.
- 4
Let the buyer shortlist on screen
Group the strongest designs into a collection and send the buyer a share link. They can approve, request a change or comment on each design without creating an account.
- 5
Stitch physical samples only for the shortlist
Make proto and fit samples only for the designs the buyer approved. Keep lab dips, PP and shipment samples exactly as your buyer requires.
Who benefits most?
Mills and fabric converters that sell cloth to garment makers can show each fabric as finished garments rather than swatches, which helps buyers picture the end product. See AI for textile factories.
Garment exporters that develop ranges for overseas buyers can shorten the back-and-forth before the first physical proto.
Smaller brands that cannot justify a 3D team can get the front-end benefit, faster design decisions, with AI previews alone. The trade-offs are compared in AI preview vs physical sample.