How to photograph garments for AI virtual try-on accurately

A reliable garment-reference workflow

For a Dreamina reference-led outfit concept, photograph the real garment before editing: a complete front view, back view and clear detail photos, plus an on-person view when drape matters. Keep the originals and measurements together so you can check every generated version against the same item.

Work through five decisions:

1

1. Choose a flat-lay, mannequin or on-person presentation suited to the intended workflow.

2

2. Capture the entire garment with controlled light and undistorted proportions.

3

3. Add the views that explain its pattern, construction and intended wearing position.

4

4. Prepare the person image and clean upload copies separately.

5

5. Compare the generated garment with the originals before creating more poses or scenes.

If you are still choosing a downstream workflow, the guide to virtual clothing try-on tools helps distinguish creative outfit images from dedicated product-to-model routes. Keep the photography set useful for either choice.

Separate pattern accuracy, visual drape and physical fit

Pattern accuracy concerns the actual print, texture and construction. Visual drape concerns how a generated garment appears to hang on a pictured person. Physical fit concerns whether a real size fits a real body; it needs measurements and physical evidence beyond a generated image.

These are different review jobs. A dress may retain its floral print but sit too high at the waist. A jacket may hang plausibly while losing a pocket. A visually flattering outfit says little about whether the sleeve or chest measurement suits the shopper.

Set the intended use before shooting. An outfit mood board can tolerate creative interpretation that a product listing cannot. For merchandise, keep the original photographs, size information and actual construction as the reference for approval.

1. Choose the right way to present the garment

Choose the presentation your intended try-on route accepts and the information the garment needs to show. Flat-lay, mannequin and on-person photos solve different problems; none makes missing construction visible automatically.

For a patterned blouse, a clear flat-lay can explain the print while an on-person view explains how the fabric falls. For a structured jacket, a mannequin may reveal shape that disappears when laid flat. Use the presentation that resolves the ambiguity rather than choosing one format for an entire catalog.

If the next deliverable is an invisible-body product image, photograph the inner neckline and other exposed interior areas too. The workflow for creating ghost mannequin images from flat-lay clothing addresses that separate production step.

2. Prepare the garment without changing its cut

Remove lint and accidental packing creases, arrange sleeves and hems, and show the garment in its intended wearing condition. Preserve designed pleats, gathers, asymmetry and fabric texture; do not pin the waist narrower just to make the photograph look neater.

Follow the care label when steaming or pressing. A deliberate crinkle finish should remain a crinkle finish. Smooth a sleeve enough to reveal its outline without pulling a stretch knit into a longer, narrower shape.

For the main reference, make collars, cuffs, hems and closures visible. Keep the intended buttoning or fastening consistent across views. Capture an additional open view when a lining or inner fastening matters, instead of combining incompatible states in one ambiguous photograph.

Remove styling objects from the garment area. A belt that belongs to the product needs its own clear view; a styling belt that changes the silhouette should not silently become part of the item.

3. Keep the camera square to the garment

Keep the camera parallel to the main garment surface and far enough away to include its whole outline. Check that the neckline, sleeves and hem look proportionate before taking close-ups.

For a flat-lay, photograph from directly above rather than from the near edge of a table. For an upright garment, avoid pointing steeply down from head height or upward from the floor. Those angles can make one end look wider and change the apparent length.

Use a stable support and leave space around the outline, including cuffs and straps. Avoid framing so tightly that a small rotation clips a sleeve. Move back if perspective looks exaggerated; a wide camera view taken very close to the garment is often less useful than a more distant, straight view.

A quick symmetry check helps reveal camera tilt, but do not force an intentionally asymmetric design into symmetry. Compare the photograph with the real garment before deciding that an uneven hem needs correction.

4. Build a small, repeatable lighting setup

Start with a stable camera or phone, a neutral background, a large soft light and a white reflector. Add a second matched soft light when one side of the garment remains dark; repeatable coverage matters more than buying the brightest lamp.

A useful basic kit includes a stable support, a clean background, one or two controllable diffused lights, secure stands and a reflector. Use clips to hold the backdrop or reflector, rather than reshaping the garment's waist. Choose equipment that fits the room and can be left in marked positions.

Keep some distance between an upright garment and the backdrop when space allows. This gives you more control over shadows and edge separation. A white garment can be easier to inspect against light gray than against an overexposed white background.

Think of the light roles separately: the main light reveals form, fill opens dark folds, and an optional background or edge light provides separation. Add the third source only when it solves a visible problem. More lamps also introduce more reflections and adjustments.

Match the light to flat-lay, mannequin or on-person photos

For flat-lays, spread soft light across the full garment from above or balanced sides. For upright garments, two diffused lights placed to either side are a useful starting point; adjust their height and angle until sleeves, torso and hem remain legible.

Start with the garment in position and look at the darkest area, not only the middle of the shirt. If a long sleeve falls outside the soft light, widening the coverage helps more than brightening the already well-lit chest. Check both shoulders and the lower hem after each change.

Around 45 degrees to either side is a practical starting arrangement for upright garments, not a fixed requirement. Raise or turn the sources to reduce collar shadows and reflections, then examine the full item again.

In a small room, use a white reflector to lift the shadow side before adding another lamp. A separate background light is useful only when the backdrop needs it: turning up the front lights until white fabric loses its seams defeats the purpose. Mark successful light and camera positions so the next colorway can be photographed consistently.

Handle black, white, shiny and sheer fabrics deliberately

Change the lighting when the fabric loses information: preserve seams in white fabric, folds in black fabric and controlled reflections in satin. Use background contrast to show transparent areas without inventing extra opacity or transparency.

• White fabric: check that stitching, lace and surface detail remain visible in bright areas. A pure white patch with no texture is a reason to adjust exposure or light.

• Black fabric: add gentle fill or separation so the sleeve and body do not merge into one shape. Keep black looking black rather than lifting it into gray.

• Satin, sequins and coated fabrics: change the source angle and apparent size to control hotspots. Inspect whether a reflection hides a seam or printed detail.

• Sheer fabric: show the real material against a suitable contrasting surface or its intended underlying layer. Keep a reference that explains what is actually transparent.

For sheer or heavily textured items, take both a full view and a detail. A dramatic backlit image can be useful styling, but it should not be the only record of the product's color and coverage.

5. Lock color and exposure for the session

Use one consistent lighting environment and a fixed or custom white balance where your camera permits it. Check a neutral reference at the start and after lighting changes, then compare the photograph with the actual fabric rather than correcting each image by eye independently.

Avoid combining changing window light with a differently colored lamp. If the light changes midway through a batch, make a new reference capture and recheck the affected garments. The aim is a consistent relationship between the light, camera and fabric.

Keep a neutral reference frame separately from the clean garment input. It helps assess whether cream has become yellow or navy has shifted toward purple. Do not warm or cool a garment simply to make it look more attractive, especially when the same style has closely related colorways.

View the images on a reasonably consistent display, but remember that a screen preview is not a color measurement. Preserve the original capture so later edits can be traced back to the same starting point.

Choose camera settings that keep the whole garment sharp

Use a low ISO when the available light allows, enough depth of field for the garment and a shutter speed that avoids blur. The right exposure depends on light, lens, distance and movement; inspect texture and seams at full size before photographing the batch.

On a camera with manual controls, ISO 100–400 and roughly f/5.6–f/11 can be starting points under adequate continuous lighting. Adjust for the lens, working distance and how far the sleeves or hem sit from the focus plane. These are photography examples, not specifications for an AI model.

A tripod allows a slower shutter for a still garment, while a moving person needs a speed that keeps the fabric sharp. With flash, stay within the camera's flash-sync limit and check the resulting exposure rather than copying continuous-light settings blindly.

Continuous lights make the effect easy to see while arranging the item. Flash is also useful when set up correctly. Whichever you use, check one complete image at full size before assuming the rest of the batch is sharp.

6. Capture a usable set of views

Keep a complete front and back view, then add the side, construction and detail views needed to explain this particular garment. A detail photograph supplements the full silhouette; it does not replace it.

Use this as a shot list, adjusting it to the item:

Check the entire outline before moving to details. Unfold a cuff that hides the sleeve length, remove a hand from a pocket you need to show, and separate overlapping straps. Photograph a hidden construction feature directly rather than hoping a close-up of a nearby area will explain it.

For sets, keep individual garment photographs as well as the combined outfit. This makes it easier to tell which top, bottom or accessory a later instruction refers to.

7. Give repeating patterns a clear scale and direction

Photograph enough of the garment to show several complete pattern repeats and their relationship to the seams. Add a sharp close-up, keeping its identity tied to the same garment and colorway.

For checks, include the placket or center seam so the viewer can see whether the pattern actually matches across it. For a floral print, show several whole motifs and their spacing across the body. A close-up of one flower cannot tell the generator how large that flower is relative to the whole dress.

Keep a separate measurement photograph with a ruler if repeat size is important, then record the measured distance in the product notes. Use an uncluttered garment photo as the main generation input unless the selected route asks for a measurement image. The ruler is a reference for your verification, rather than a promise that the model will calculate exact print scale.

Also check sharpness at the pattern's smallest meaningful detail. Fine checks and tightly woven textures can produce false-looking bands or color in a photograph. If that happens, change the camera distance slightly and inspect a new capture. Do not pass an already distorted reference downstream and expect an AI edit to recover the original weave.

8. Record dimensions separately from the image

Record the actual garment size and relevant measurements in a separate product record. Distinguish flat measurements from circumference and note stretch or intended ease; a visual try-on cannot supply missing measurements reliably.

For example, record whether “waist 36 cm” means the width of the garment laid flat, and specify how it was measured. Do not mix a relaxed waistband measurement with a stretched measurement without labeling them. Keep the garment's size label and the seller's measurement method with the same SKU.

Choose measurements that matter to the design: body and sleeve length for a shirt; waist, rise and inseam for trousers; waist position and length for a skirt. Include fabric composition or stretch information where available. Do not derive those facts from a generated image.

An on-person reference can show how that particular sample hangs on that person. For a sizing decision, add the appropriate real measurements and fit information rather than transferring that appearance automatically to another body.

9. Prepare a separate, unobstructed person photo

Use a clear person image with the intended garment area visible and a simple pose. Move hands, hair and accessories away from important clothing regions, and use an image you have permission to upload.

A front-facing pose with arms slightly clear of the torso often makes a top easier to inspect. Keep the full relevant area in frame: a cropped waist is a poor starting point for judging skirt placement, and missing feet can make trouser length harder to assess.

Avoid poses that conceal the product behind crossed arms or a bag. For a jacket, decide whether you need an open or closed appearance. If a belt, coat or long hair will hide the defining detail, first create a clearer view before adding the styling.

The garment photo and person photo have different jobs. One documents the product; the other supplies the person and pose. Label them clearly rather than relying on the order of an unlabeled collage.

10. Prepare upload copies without losing the originals

Retain the original high-quality photographs, then make clean upload copies that meet the selected tool's actual format and size requirements. Avoid repeatedly downloading, resizing and recompressing the same reference.

Check the intended entry's requirements before exporting. For example, Google Merchant Center's apparel-try-on guidance specifies images of at least 512 × 512 pixels and prefers 1024 pixels or higher. That is a requirement for that Google route, rather than a universal minimum for Dreamina or every try-on model.

Keep the whole garment in the upload copy. Correct rotation and a slight crop around unused background if needed, but avoid stretching width and height independently. Do not sharpen a tiny logo until it becomes a different logo or use generative enlargement to invent a print detail.

If you clean the background, compare the result with the original around hems, lace, straps and sheer panels. A tidy background is useful only if the garment remains a trustworthy reference. Keep measurement frames, labels and review notes in the archive even when they are not suitable primary generation inputs.

Use the photographs in Dreamina

In Dreamina, choose an image workflow that accepts the references you need, upload the actual garment and person references where supported, and describe one clear visual change. Keep the garment's defining features explicit and compare the result with the originals before building a new scene.

For example, a useful instruction identifies the garment and what makes it recognizable: the navy-and-cream stripe direction, the waistband position and the side opening. Ask for the desired outfit or pose without simultaneously requesting a new fabric, tighter cut and different print. That keeps the creative direction easier to judge.

Treat a prompt such as “keep the garment unchanged” as an instruction to inspect, not a preservation lock. If a buckle disappears or a stripe changes direction, return to the clearer product reference or make a narrower edit. Confirm the commercial-use permission that applies to your market, selected model and plan before using the result in a listing or advertisement.

Once the garment is acceptable, develop the scene around that approved direction. Keep a clean product reference available throughout the process so a pleasing new background does not distract from a changed cuff or hem.

A concrete downstream example: FASHN Agent

FASHN Agent's current guide separates a model image from a product image and accepts on-person, flat-lay and mannequin garment photos. Its preference for on-person input is useful when visible drape matters; it is not a universal rule for every try-on tool.

Attach the two images and explicitly request the try-on. For several items, keep each garment's role clear; the guide allows up to six images in a message. This upload limit is separate from the larger photo archive you keep for verification.

The guide's skirt example initially places a low-waist design too high, then corrects it through a follow-up instruction. Use that as a reminder to check wearing position against the product reference. After a correction, compare the waistband and nearby pattern again before continuing.

11. Check the generated garment before making more images

Compare the generated garment with the real front, back and detail photographs before approving a pose, scene or campaign variation. Start with the full silhouette, then inspect print, seams, trims and the intended wearing position.

Look at the whole image and then enlarge the important regions. A plausible overall impression can hide a reversed pattern on one sleeve or a buckle that has become decorative stitching.

For a layered jacket, check whether it sits over the intended inner top and whether its fastening state matches the brief. For a top-only change, inspect the surrounding outfit as well: an attractive result may have altered something you meant to retain.

Approve the garment before asking for several new scenes. Otherwise, a small error in the first image becomes a repeated design feature across the entire set.

A worked example: a striped low-waist skirt

For a striped low-waist skirt, capture the full front and back, a repeat-scale detail and the waistband construction, then record the actual size and measurements. Use the intended low-waist position as a visual instruction and verify it in the output.

Suppose the skirt has navy-and-cream vertical stripes, a side opening and a deliberately asymmetric hem. Build a reference pack with:

• A straight front view showing the complete waistband and hem.

• A real back view showing how the stripes and closure continue.

• A detail photograph that includes several stripes and the side seam.

• An on-person view that clearly shows the intended low-waist placement.

• Separate measurement notes identifying the actual size and measurement method.

A useful proposed instruction is: “Show the person in the supplied navy-and-cream striped skirt, worn low on the hips as in the wearing reference. Retain the vertical stripe direction, side opening and asymmetric hem. Use a simple standing pose with the waistband visible.”

Then compare the output in the same order. Did it raise the waistband? Narrow the stripes? Close the side opening? Straighten the asymmetric hem? Correct the particular problem or improve the reference that explains it.

Fix the input that caused the problem

When a result changes the product, identify the lost information before rewriting the entire prompt. Some problems need a clearer photograph or a different input; others need a narrower edit or a new capture.

Work on one cause at a time. A brighter, sharper, more saturated edit can make the image look better while making the garment less accurate. When the source lacks real information, return to the item rather than stacking more creative instructions onto the same weak reference.

Keep a small reference pack for every colorway

Give each SKU and colorway its own originals, upload copies, measurements and approved visual checks. Record the lighting and camera positions so the next item starts from a consistent setup.

Use descriptive filenames such as skirt_navy-cream_front_original and skirt_navy-cream_waist-detail. Keep an untouched original separate from the cleaned upload file. Record which garment size was photographed and which wearing view supplied the styling direction.

Before finishing a session, check the shot list while the item and lighting are still available. Reshooting a missing fastening or blurred label now is easier than rebuilding the setup after generation exposes the gap.

If these photographs will also feed your store's product listings, choose the main and supporting images for that destination separately. The guide to Google Shopping product images explains how that delivery decision differs from assembling try-on references.

Frequently asked questions

Choose the lightest capture setup that records this garment clearly, then use the extra views to resolve specific ambiguities.

Can I use a phone or natural light?

A phone and soft window light can produce useful garment references when the entire item is sharp, evenly lit and correctly proportioned. Keep the session short enough for the light to remain consistent and recheck color when conditions change.

Use stable support, avoid harsh direct sun and inspect the actual saved photograph. Move the garment or reflector if one sleeve falls into shadow. If the window light changes too much between products, a controlled light can make the next batch easier to repeat.

Is one front photo enough?

One clear front photo may be enough to start an accepted try-on route, but keep back and detail photographs for verification. The number of images you photograph is separate from the number a particular generator accepts in one request.

Prioritize the main image that communicates the garment best. Add the most relevant detail or wearing reference when supported, and keep the remaining views for checking the result rather than forcing them into a busy collage.

Can AI restore an unseen back or hidden print?

An AI tool can propose a plausible unseen surface, but it cannot verify the real construction from a front-only photograph. Capture the missing angle before using the generated image to describe an actual product.

The same applies to a hidden label, lining or fastening. A creative reconstruction may suit a concept image; merchandise needs the actual feature.

Does a convincing try-on tell me which size fits?

A convincing image can communicate a styling idea, but it does not establish physical fit for the viewer. Keep the size chart, actual garment measurements and relevant fit information alongside the visual.

For decisions that depend on tightness, movement or stretch, use the appropriate real fitting evidence. A generated view is useful for appearance, while a sizing recommendation needs its own support.

Start with a photograph you can trust

Start your Dreamina visual workflow with a complete, sharp and honestly styled garment reference. Clear capture makes the next creative decision easier, while the retained originals make changes to the actual product easier to catch.

Choose one garment, capture the missing views and review a first outfit concept before scaling to a whole collection. Build more scenes from an approved garment direction, with the real item still available as your reference.