Drape
Beta
AI Fashion

How To Photograph Clothes For AI Try-On

The garment photo determines half your result quality. A practical checklist for shooting product images that survive segmentation cleanly — usable by anyone with a phone.

By Drape Editorial
Last updated June 11, 2026
5 min read
Quick answer

How do I photograph clothes for AI virtual try-on?

For AI virtual try-on, photograph the garment as a flat-lay against a plain white or light-grey background, lit evenly with soft natural daylight. The garment should fill 70–90 percent of the frame, be photographed from directly above with the camera parallel to the surface, and include every detail — collar, cuffs, hem, buttons. Avoid wrinkles, folds that obscure the silhouette, hangers visible in the frame, watermarks, and price tags. Use at least 1024 pixels on the long side and save as JPG or PNG.

Key takeaways
  • Shoot flat-lay against plain white or light grey.
  • Soft natural daylight, no harsh shadows.
  • Garment fills 70–90 percent of the frame.
  • Camera directly above, parallel to surface.
  • No hangers, watermarks, price tags, or visible labels.
  • Minimum 1024 px on long side, save as JPG/PNG.

The flat-lay setup that works every time

A flat-lay is a photograph taken with the garment laid flat on a surface and the camera directly above. It is the format AI virtual try-on systems were trained on, and it is also the easiest setup to achieve at home.

Start with a plain background — a white bedsheet, a sheet of cardboard, a light wooden floor. Avoid patterned fabric, glossy surfaces (which reflect the camera), and very dark backgrounds (which fight the segmentation step).

Lay the garment flat. For tops and dresses, button or zip it fully, smooth the wrinkles, and arrange the sleeves at a natural slight angle. For bottoms, lay them flat with one leg slightly forward. For suits and blazers, button the front and arrange the lapels so both sides are visible.

Shoot from directly above. If you can stand on a chair or table without endangering yourself, do so. A phone held at chest height pointing down produces a perspective-distorted image that the segmentation network handles imperfectly.

Lighting that beats studio gear

You do not need softboxes or ring lights. Natural daylight from a north-facing window (in the northern hemisphere) is the gold standard — soft, diffused, and even. Avoid direct sunlight, which creates harsh shadows.

If you must shoot indoors at night, place two desk lamps with white shades on opposite sides of the garment at 45-degree angles. The goal is to eliminate any sharp shadow lines on the fabric.

What to avoid in the frame

Five elements ruin otherwise-good product photos. Visible hangers tell the segmentation network the garment has extra geometry that does not exist. Watermarks and copyright text are interpreted as part of the design — they will appear printed on your try-on result. Price tags and security tags produce the same effect. Folded or rolled sleeves cause the model to render only the visible cuff, ignoring the rest of the sleeve. Background props (other clothing, jewelry on the same surface) bleed into the alpha mask.

Frequently asked questions

Can I use a product photo from a retailer's website?+
Yes, with two caveats. First, check the retailer's terms of use — many prohibit commercial use of their imagery. Second, retailer photos often include subtle watermarks or studio shadows that will appear in your try-on. Crop tightly to just the garment.
What about screenshots from videos?+
Single-frame screenshots usually work but are typically lower resolution (720p) than ideal. The AI may struggle with motion blur. Pause on a clean frame, screenshot, and crop.
Do I need a DSLR?+
No. A modern phone camera (any iPhone 12+ or Pixel 6+) produces images well within the quality range that AI try-on needs. Resolution is more important than sensor size.
Drape Editorial
AI Fashion Research

Drape Editorial is the in-house research team behind Drape Try-On. We test virtual try-on models against real garment photography weekly and publish what we learn.