1. Shoes fit differently on different people

Models are created to a specific size and adjusted for VTO, but perfect alignment isn’t always possible, especially where there is significant variation in foot size. It’s expected that the model may not look perfect on very large or very small sizes. It is optimised to work well in general.

Recommendation: Test women’s models on women’s feet and men’s on men’s.

2. Poor VTO from the back of the foot

Image jump/unstable tracking from rear angles.

This viewpoint is uncommon for product evaluation, so the behaviour does not materially affect the try-on experience or how the shoe is perceived.

3. Low-contrast background issues

When trying on shoes against a low-contrast background, foot detection may be imperfect. As a result, shoes can appear larger or smaller than the actual feet, or may fail to align with the feet altogether. 

This is a known limitation; long-term improvements are planned. For now, this behaviour is expected when using challenging backgrounds.

4. Floor visible through the model

The floor may be visible through shoe models at certain viewing angles, occurring mainly on the Web and less frequently on iOS. 

This is a known limitation; long-term improvements are planned. For now, this behaviour is expected when trying on from certain angles.

5. Flor visible between toes (open-toe footwear)

For sandals, flip-flops and other open-toe styles, the floor may be visible between the toes, and in some views the toes may be partially hidden by the shoe sole.

This is a known limitation; long-term improvements are planned. For now, this behaviour is expected.

6. Shoe tip raised in VTO (high heels only)

For high-heeled styles (around 7 cm), the shoe tip may appear slightly raised during VTO. 

This is an intentional trade-off to ensure the heel looks realistic in side views, which are more important for how users evaluate and perceive high heels. As a result, side-view fidelity is prioritised over perfect toe alignment in other angles. When viewed from the side, this issue does not occur.

7. Side-view occlusion issue (high boots only)

For high boots, the occluder may not work correctly in side view. When the side of one shoe overlaps the collar of the other, a hole can appear along the side.

1. Fitting issues

At certain angles, the bag may appear to dig into the shoulder, and the strap can look unnaturally curved over the shoulder.

These are known limitations; long-term improvements are planned. For now, this behaviour is expected.

2. Forearm visibility required for proper alignment

For proper alignment and occlusion, the user’s forearm should be at least partially visible in the camera view. If the forearms are out of frame - especially at certain angles - occlusion may fail and some “holes” (occlusion artefacts) can appear on the bag.

This is a known limitation; long-term improvements are planned. For now, this behaviour is expected.

3. Camera tilting distorts alignment

For accurate alignment, keep the camera in a vertical (portrait) position. Rotating or tilting the device can disrupt alignment and lead to rendering artefacts and distorted bag proportions.

This is a known limitation; long-term improvements are planned. For now, this behaviour is expected.

4. Perceived bag size may vary

Proportions may appear slightly larger or smaller than the actual product as they adapt to the human body. As a result, strap length can sometimes look unrealistic as well. Providing the user’s height improves accuracy, but only when the entire body is visible in the frame.

1. Gaps between the scarf and the user’s body

Small gaps may appear between the scarf and the user’s neck, hands, or other body parts when viewed from certain angles. 

2. Occluder mismatch with the user’s body shape

The occluder (the virtual body model used for occlusion) may not precisely match the user’s shape, which can lead to minor visual inconsistencies around the scarf - e.g., along the neck, shoulders, or arms. 

3. Occasional holes on the scarf surface

For proper alignment and occlusion, the user’s head and hands should be at least partially visible in the camera view. Fully occluding either can cause rendering errors.

Additionally, specific folds or high-contrast patterns on clothing can confuse the detection algorithm, leading to incorrect boundary detection around the neck, arms, or head. Where possible, use plain or low-contrast clothing.

4. Camera tilting distorts alignment

For accurate alignment, keep the camera in a vertical (portrait) position. Rotating or tilting the device can disrupt alignment and lead to rendering artefacts and distorted scarf proportions.

5. Scarf length varies by user height

Despite 3D models of long scarves are generated to real-world dimensions, the displayed length can vary with the user’s height. As a result, the scarf may appear slightly longer or shorter than in real life when worn on body. This is a known limitation of on-body scaling.