Apparel
Flat-Lay to On-Model Photos: AI Virtual Try-On for Apparel Sellers
On-model photos sell apparel better than flat-lays for most categories, but booking models for every colourway is the most expensive line in a fashion seller's content budget. Virtual try-on — generating an on-model image from a flat-lay or ghost-mannequin shot — is now good enough for secondary images and many marketplaces, provided the garment stays accurate.
What "accurate" means for apparel
- Print and pattern: stripes line up, logos are not re-drawn, all-over prints keep their scale.
- Colour: the exact shade, not a "close" one — colour mismatch is a top return reason.
- Construction: neckline, pocket count, button count, hem length.
- Fabric behaviour: knit drapes differently from denim; the image should not make a stiff fabric look fluid.
Prepare the references
- Shoot the flat-lay straight down, evenly lit, on a neutral background, garment smoothed.
- Add a close-up of any print or label so the model has detail to copy.
- Pick or build a consistent model reference with character sheet so all your colourways use the same person.
- Choose a pose with pose reference that shows the garment's key features — front, hands away from pockets.
Generate
Run the try-on on outfit change. Flux Kontext and Flux 2's multi-reference input are the strongest at keeping the garment and the person consistent at once.
Dress the model in the exact garment from the reference: same colour, print scale and placement, neckline, sleeve length and button count. Natural fit for a size M, studio softbox light, light grey seamless background, full body, front view. Do not change the model's face or hair.
Check for drift
Place the generated image next to the flat-lay and check, in order: colour (sample both), print placement, logos, construction details, length. If one detail is wrong, fix it with a masked inpaint pass rather than regenerating the whole image.
From still to motion
A short walk or turn sells fit better than a still. Keep the same model across clips with consistent character video.
Cost per colourway
At list price one try-on image is $0.038 on Flux Kontext dev or $0.045 on Flux 2 Pro edit. Four on-model angles per colourway with a retry each stays in the low single dollars.
常見問題
Is AI virtual try-on accurate enough to list?
For colour, silhouette and print it often is, but always compare against the real garment. Fix wrong details with a masked edit rather than accepting drift.
Can I use the same model across all my products?
Yes. Build a character sheet once and use it as a reference for every try-on so the face and body stay consistent.
Ghost mannequin or flat-lay — which input is better?
Ghost-mannequin shots give the model shape information and often produce better fit. Flat-lays work well for prints and simple cuts.
What does one try-on image cost?
$0.038 on Flux Kontext dev at the current list price, per output image.