How to Photograph Flat-Lay Garments for AI Try-On Tools

The single biggest driver of AI try-on output quality is the input photo, not the tool. Across our test runs, the same generator produced usable output from a well-shot flat-lay and unusable output from a wrinkled, angled, mixed-lighting shot of the same garment. Fixing the input is cheaper than upgrading the subscription.

Shoot straight down, not at an angle

Garment-warping models estimate how fabric should drape from a flat reference. A photo shot at an angle gives the model a distorted shape to start from, and it propagates that distortion into the output. Mount the camera or phone directly above the garment, parallel to the surface.

Flat, even lighting beats dramatic lighting

Two soft light sources at roughly 45 degrees on either side, or an overcast window, eliminate the hard shadows that a model will read as fabric features. A shadow across a sleeve frequently comes back as a crease or a colour shift in the generated image.

Iron everything, including the parts you think will not show

Wrinkles are the most common cause of what looks like model failure. A crease in the source is a crease in the output, often exaggerated. In our runs, ironed garments had a noticeably higher keeper rate than unironed ones on the same tool.

Use a plain background with contrast

White garment on white background is the hardest case for segmentation. The model has to guess where the garment edge is, and it guesses wrong at the hem and cuffs. Use a mid-grey surface for light garments and a light surface for dark ones.

Shoot larger than you need

Amazon requires a 1600px minimum and recommends 2000px. Shopify recommends 2048x2048 and supports up to 5000x5000. Generation tools downsample your input, so starting large gives the model more detail to preserve. Shooting at your minimum output size guarantees a soft result.

Keep the garment shape natural

Do not stretch sleeves flat into a T shape if the garment does not hang that way. Let the shoulders sit at their natural width and arrange sleeves with a slight, natural bend. Over-flattening produces output where the fit reads wrong even though the garment is technically correct.

What to test before committing to a tool

Pick your two hardest garments — usually something with a dense repeating pattern and something sheer or layered — and run both through any tool's free tier before you subscribe. Solid-colour garments look good on every tool in the category; the hard cases are where the difference is. Our published sub-scores for garment fidelity are built on exactly this principle.

Frequently asked questions

Does a mannequin shot work better than a flat-lay?

For most tools yes, because a mannequin already provides three-dimensional drape information. Ghost mannequin shots in particular tend to produce better output than pure flat-lays. Check which input types your chosen tool accepts before reshooting.

What resolution should I shoot at?

At least 2000px on the long edge, and larger if your camera allows. Generation tools downsample, so extra input detail improves the result even when the output is smaller.

Why does my patterned garment come out distorted?

Dense repeating patterns are the hardest case for garment-warping models. Shoot the pattern square to the camera with even lighting, and expect to test several tools — this is the dimension where scores diverge most.