Why AR shopping changes how you shoot
Augmented reality and interactive 3D viewers are no longer a novelty reserved for furniture giants. Shoppers who view a product in AR are roughly 65% more likely to buy, and listings backed by 3D visualization see about 22% lower return rates than those with static images alone. That combination — more conversions and fewer returns — is why AR product photography has moved from "nice to have" to a planning line item for serious catalogs.
The catch: AR and 3D pipelines are unforgiving about source imagery in ways flat product shots never were. A 2D hero image only has to look good from one angle. An AR-ready asset has to hold up when a customer spins it, tilts their phone, and views it against their own living room or their own wrist. Gaps in coverage, inconsistent lighting between angles, and baked-in shadows that don't match the real world all break the illusion instantly.
Two paths: photogrammetry vs. AR-augmented 2D
There are two practical ways to get a product into an AR experience, and the photography requirements differ sharply between them.
Photogrammetry / 3D capture
- Shoot 40–120 overlapping frames around the product
- Software reconstructs a true 3D mesh and texture
- Best for rigid, opaque objects: shoes, bottles, electronics
- Struggles with glass, chrome, fur, and soft fabric
AR-augmented 2D
- A single clean cutout placed as a flat or billboarded plane in the user's space
- No mesh required — fast and cheap to produce
- Best for wall art, rugs, mirrors, signage, packaging mockups
- Looks flat the moment the viewer walks around it
Most catalogs use both: photogrammetry for hero SKUs that benefit from true rotation, and AR-augmented 2D for flat or low-relief products where a plane is convincing enough. Decide per category before you shoot — the capture setup is completely different.
The turntable setup that actually works
For 3D capture, consistency between frames matters more than the beauty of any single frame. Your goal is identical lighting and exposure across every angle so the reconstruction software can match features cleanly.
- Lock everything manual. Fixed aperture, shutter, ISO, white balance, and focus. Any auto setting that drifts between frames introduces seams in the final texture.
- Light the product, not the background. Diffuse, wraparound lighting from multiple sides minimizes hard shadows that "stick" to the model as it rotates. Shadows that move with the object look wrong in AR.
- Use a marked turntable. Equal angular steps — every 9° for 40 frames, or every 5° for 72 — give the software even coverage. Eyeballing the rotation creates dense and sparse zones.
- Shoot multiple elevations. One ring at eye level isn't enough. Add a high ring (~30°) and a low ring to capture the top and underside, or the mesh will have holes.
- Keep the camera fixed, rotate the product. Moving the product on a turntable while the camera stays put is far more repeatable than walking the camera around.
Add a few physical markers or a coded target mat under the product. They give photogrammetry software high-contrast reference points to lock onto, dramatically improving reconstruction on smooth or low-texture surfaces.
Lighting and shadows for a believable AR placement
The single biggest tell that an AR product is fake is lighting that doesn't match the room. A bottle lit with a hard left-side highlight will look pasted-in when a shopper places it on a sunlit counter.
The fix is to capture your product under flat, neutral, even light so the AR engine can relight it to match the user's environment. Bake as little directional lighting and shadow into the texture as possible. The same principle applies to 2D cutouts destined for AR: a soft contact shadow that the AR renderer can fade or replace beats a dramatic baked-in shadow that fights the real floor.
| Lighting choice | Static hero image | AR / 3D asset |
|---|---|---|
| Hard directional key light | Great for drama | Breaks relighting |
| Soft, even wraparound | Can look flat | Ideal |
| Baked-in cast shadow | Grounds the product | Fights the room |
| Neutral white balance | Accurate color | Essential |
Handling the hard materials: glass, metal, and fabric
Photogrammetry reconstructs surfaces by matching features between frames. Materials that don't hold a stable appearance from angle to angle — anything transparent, mirror-finished, or moving — are where pipelines fall apart.
For chrome and glass, cross-polarization (a polarizing filter on both lights and lens) cuts the moving reflections that confuse reconstruction. For fabric and fur, you often can't capture a clean mesh at all — these are the SKUs where AI-assisted cleanup or a hybrid approach earns its keep. Tools like Retouchable can generate consistent multi-angle product views and clean cutouts from a small set of source frames, filling the gaps where a raw photogrammetry pass leaves holes or noise on difficult materials.
Exporting and testing before you ship
Capturing the frames is half the job. The asset has to survive compression, load fast on a phone, and look right in the actual viewer.
- Target glTF/GLB for web AR. It is the standard for both iOS Quick Look (via USDZ conversion) and Android Scene Viewer. Keep meshes under a few hundred thousand polygons and textures at 2K where possible.
- Test on real devices, not just the desktop preview. Lighting, scale, and load time behave differently on a mid-range phone over cellular than on your workstation.
- Verify real-world scale. An AR sofa that appears the size of a shoebox erodes trust instantly. Set physical dimensions explicitly in the model metadata.
- Check the contact shadow. The product should look grounded, not floating. A soft auto-generated shadow under the model sells the placement.
Don't reuse your heavily-retouched 2D hero shot as an AR texture. Aggressive dodge-and-burn and baked highlights that look stunning flat will read as dirty smudges once the model rotates under new lighting.