AI Virtual Try On Shoes
The AI virtual shoe try on replaces the footwear in a photo with a pair of your own. Upload a picture of a model, or of yourself, then an image of the new shoe design, and the AI puts the selected pair in place of the original one while keeping proportions and lighting realistic. Built on computer vision and machine learning, it saves brands and designers a lot of the time that footwear visuals usually cost.
- 1Upload your model feet photo
- 2Upload your new shoes image
- 3Try on different shoes with AI
This demo was recorded in our previous interface — an updated video of the new studio is on its way.
How it works
Upload your model feet photo
Start with a photo where the feet are visible. This is the image the new shoes will be placed into.
Upload your new shoes image
Then add the picture of the pair you want to see worn. The AI swaps it into the original photo, on the same feet, in the same pose.
Try on different shoes with AI
Every type of shoe can go through the same pass, so one model photo can carry a full range. Change the pair, run it again, and the series stays consistent.
Can you change the shoes in a photo for free?
You can change the first pairs on the credits an account opens with, and after that each swap costs one credit on Standard or three on Pro. Uploading costs nothing, and neither does looking. The meter runs on the generation, so putting six pairs on the same photograph costs six.
It is worth being clear about what "try on" means here, because the phrase covers two different things. This is a photographic swap: the footwear in an existing picture is replaced by a pair you supply, keeping the pose, the light and the rest of the image. It is not a fitting: nothing measures a foot, and nothing tells you whether the shoe fits the person in the photograph. What it answers is how the pair looks worn, which is the question a product page and a lookbook are actually asking.
What makes a photo swap cleanly
Feet are small, near the bottom of the frame, and often partly hidden, which makes the source photograph matter more here than in a full-garment swap.
Four things help. Feet fully in frame, with a little ground around them, so the shadow the shoe casts has somewhere to land. A straight-on or three-quarter angle rather than a steep overhead, which flattens the shoe's profile and hides exactly the part that identifies it. Both feet visible if the pair is the subject: a hidden second shoe is invented, and invented shoes rarely match. And a still pose, because motion blur at the ankle gives the model nothing solid to attach to.
The pair you upload matters too. A clean product shot of the shoe, side or three-quarter, on a plain background, transfers far better than a styled photograph where the shoe is already worn by somebody else.
Where it struggles, and how fast you find out
Two honest limits. Extreme angles, a foot pointing directly at the camera or a shoe seen from behind, are where the swap loses the silhouette, and it is often quicker to reshoot the base photograph than to try again. And very tall boots interact with the trousers or the hem above them, so a knee boot under a long coat is a harder ask than a sneaker with a cropped jean.
Frequently asked questions
What does the starting photo need?+
Visible feet. Unlike the rest of try-on, this workflow is a replacement: the AI takes the pair already in the photo and swaps in yours, on the same feet, in the same pose, with proportions and lighting kept realistic. The feet are the anchor everything else is built on.
Does the pose or the rest of the photo change?+
No. The swap touches the footwear and nothing else. The model, the pose, the ground, and the light stay as photographed, which is why a series of swaps on one photo stays visually consistent.
What shoe styles can I swap in?+
Every type of shoe can go through the same pass: sneakers, heels, boots, platforms. Upload the image of the pair you want to see worn and the AI puts it in place of the original one.
How do I cover a whole footwear range?+
One model photo can carry the full range: change the pair, run it again, and the series stays consistent. For an entire catalog in one go, use bulk generation; for accessories, the same idea exists as AI virtual try on jewelry.