A 3D scan is only as good as its photos. Follow these simple rules and your first scan will come out clean — any phone or camera works.
If you remember nothing else, remember these.
Shoot from every side of the object. Anything the camera never sees cannot be reconstructed.
Each photo should share about 60–80% of its view with the previous one. Small steps, many photos.
Blur is the number-one scan killer. Hold still for each shot and shoot in good light.
The object and everything around it must stay perfectly still for the whole set.
Move around the object in small steps — roughly every 10–20 degrees — keeping it centered and always pointing the camera at it.
Two neighboring photos must show mostly the same thing. If two consecutive shots have little in common, the reconstruction breaks apart there.
Do two or three full loops at different heights — one looking slightly down, one at eye level, one from low up — so the top and undersides are covered too.
The object should take up a good half of the picture. Step closer instead of zooming, and add a close-up loop for detailed areas.
Don't change zoom mid-set, and keep exposure steady. If your camera app can lock focus and exposure, lock them.
20–60 photos suits most objects. Extra photos rarely hurt — missing angles always do.
Photogrammetry works by matching surface details between photos. Some subjects give it a lot to match — some give it nothing.
Shiny object you must scan anyway? A light coat of dust, chalk spray or removable matte spray gives the surface enough detail to match.
Soft, diffuse light is ideal: an overcast sky outdoors, or bright indirect room light indoors. Avoid direct sun, harsh shadows and flash — shadows get baked into the texture.
A background with some texture actually helps the cameras get located. Just make sure nothing in it moves — no people or cars passing through your shots.
Rotating the object instead of walking around it confuses the reconstruction, because the background stops matching. Move yourself, not the subject.
You can upload a video (.mp4, .mov, .mkv) instead of photos — frames are extracted automatically, and blurry or near-duplicate frames are filtered out. The same rules apply, plus:
Record at the highest resolution your camera offers — video frames start out softer than photos.
Walk around the object in a slow, steady 1–2 minute loop. Slower movement means sharper frames.
Quick swings cause motion blur and distortion. Keep the object centered and turn gently.
After the first loop, do another at a different height, just like with photos.
If a scan goes wrong, the photos usually tell you why.
Point cloud splits into disconnected pieces
A gap in coverage — consecutive photos with too little overlap
Reshoot with smaller steps between shots, or add photos bridging the gap
Holes or missing areas in the model
That area was never photographed clearly from enough angles
Add close-up shots aimed at the missing area from several directions
Texture looks smeared or melted
Blurry photos, or lighting that changed during the shoot
Remove blurry shots before uploading and reshoot in steady, even light
Only a few photos get used in the reconstruction
Surfaces too plain or too shiny — nothing for the matcher to grab
Matte the surface, add surrounding texture (newspaper, cloth), improve light
Ghosted or doubled surfaces
The object or background moved partway through the set
Reshoot in one pass keeping everything fixed; exclude photos taken after anything moved
Grab an object near you, shoot 20+ overlapping photos all the way around it, and see it as a textured 3D model — right in your browser. You can upload and see the price without signing up.