Flat-lay to try-on video

Open-source try-on models (OOTDiffusion, IDM-VTON, CatVTON) make stills and most carry research licences; nobody ships a clean video path. On varg it is two steps: a try-on still, then image-to-video with fit-check motion. Fit causes about half of apparel returns, so the video must show the real drape and say the size.

1. Clean the inputs

• Garment: flat-lay or ghost-mannequin photo, 1000 px or more (resolution is the biggest quality factor), tags and measurement labels removed (the model reads them as print). Write a garment-truth caption of only what is visible: fabric, colours, print or stripe pattern, buttons, pockets, logo text and position, hem length.
• Model: a full-body photo (head to shoes, arms relaxed, plain room), a generated one or a consenting real person. Keep one model per product line; for more shots, make a 4-view sheet (front, 45°, side, close-up) with varg-character-design.

2. Dress the model (still)

curl -s -X POST https://api.varg.ai/v2/image -H "Authorization: Bearer $VARG_API_KEY" -H "Content-Type: application/json" \
  -d '{"model":"bria_fibo_edit_1_5_virtual_try_on","files":[{"url":"MODEL_URL"},{"url":"GARMENT_URL"}],
       "prompt":"Dress the woman in the cardigan from the second image, worn open over her white t-shirt. Keep her face, hair, pose, jeans, sneakers and the room unchanged."}'

4 credits. For two garments or accessories, add up to three garment images. For prints, logos or embroidery Bria softens, try nano_banana_2_1 or gpt_image_2_5 with both images and: "preserve all visible text, graphics, prints and seams; do not add text, logos or accessories; full body, hem visible" (test notes). Compare against the garment-truth caption before animating.

3. Animate fit-check motions

One motion per 5 s clip, camera locked, full body and hem in frame, hands away from the print:

Motion	Prompt core
Turn	"turns slowly 90 degrees to show the side, pauses, turns back"
Tug and fall	"lightly tugs both front edges and lets go so it falls open naturally"
Walk to camera	"walks three steps toward the camera, fabric moves with weight"
Mirror selfie	"phone mirror selfie, shifts weight, checks the back in the mirror"
Detail	"close-up: fingers pinch the knit at the cuff, slow pan across the texture"

Fashion fit-check video, phone on a tripod, camera locked, full body and the hem always in frame.
<motion>. Hands stay away from the front panel. <fabric> moves with weight; <pattern, buttons, logo>
stay exactly the same. Natural window light, realistic iPhone footage, no text.

kling_v4 (79 credits per 5 s, 720p) held the stripes and buttons in our test; seedance_2_5 adds references (pass the garment as @image2) when Kling drifts. Cut 3–4 clips into a 15 s ad with a voiceover or music and captions.

4. Honest fit and labels

• Caption the size: "Model 170 cm, wears M". Offer 2–3 body types for bestsellers.
• Show realistic drape; don't let the model slim or reshape the body.
• AI model = AI label on TikTok and Meta; in ads reaching New York, a visible synthetic-performer disclosure (ai-ad-compliance-check). Never use a real person's likeness without consent.

Failure modes

Logo or micro-print melts in motion → slower motion, add a static close-up, or drop that shot. Hands on the garment → the hidden part is redrawn. Hem or length drifts → "full body and hem in frame". Colour flicker → name the exact colours. Face drift → one model photo per product, 4-view sheet for series.

Tested

2026-10-28: a cream cardigan with navy stripes and pearl buttons (flat-lay from nano_banana_2_1) on a full-body model photo. bria_fibo_edit_1_5_virtual_try_on dressed her in one call (4 credits), keeping face, room, jeans and the stripes; it dropped the two patch pockets. kling_v4 fit check, 5 s (79 credits): she turns to the side, turns back, tugs the cardigan open; stripes and buttons held. https://s3.varg.ai/files/194b94d0-58ff-4e8b-91de-4f1959d04013/file_1213e32adbda.mp4

Related skills

varg-character-design (a model who stays the same), varg-product-lock (garment details as references), ai-ad-compliance-check, varg-ad-variants (one per body type or colour), varg-ai.