Home / Guides / Getting results
How to Choose the Best Photo for Image-to-3D
When an image-to-3D result disappoints, the settings are rarely the cause. The photo is. The model can only work with what it is shown, and a few simple choices when you take or pick the photo make a noticeable difference.
The checklist
- One subject, fully in frame. Nothing cropped at the edges, nothing else competing for attention.
- Plain background, contrasting with the subject. A white sheet, a wall or a desk works. A gray cat on a gray sofa does not.
- Soft, even light. Near a window on an overcast day is ideal. Avoid hard shadows and bright highlights.
- Slightly above and to the side. A three-quarter view shows the front and one flank, giving the model more volume to work with than a flat front view.
- Sharp and reasonably large. The uploaded file may be up to 10 MB in PNG, JPEG or WEBP. Sharpness matters more than resolution; a few megapixels of crisp detail beats a blurry 20.
- Matte, opaque, solid. Mirrors, glass and chrome confuse depth estimation.
Why the background matters so much
The first thing the pipeline does is cut the subject out of the background (see how the process works). Cutout tools follow edges. If the subject's edge is clear, you get a clean silhouette and therefore a clean shape. If the subject blends into the background, or shadows fall across the ground beside it, bits of shadow or floor can end up as extra geometry, a common cause of flat "skirts" or lumps around the base.
If you already have a photo with a busy background, remove it yourself first in any image editor and save a PNG with a plain or transparent background. That gives you control over where the edge falls, which the automatic tool does not always get right on fine details like fur or wire.
Framing and angle
Leave a modest margin around the subject; the image is centered and rescaled anyway, so filling the frame gains nothing and cropping loses geometry. For most objects, imagine the camera at about the height of the subject's top, looking down at roughly 15 to 30 degrees, rotated 20 to 40 degrees away from straight-on. From there you see the top, the front and one side, the three surfaces that define the volume.
Avoid extreme wide-angle lenses at close range, which stretch the nearest part of the object. Stand back and zoom in, or crop.
Choosing subjects that work well
| Works well | Tricky | Usually fails |
|---|---|---|
| Toys, figurines, statues, mascots and characters | Thin structures: chair legs, plants, antennas | Transparent glass, mirrors, water |
| Animals and pets sitting still | Very glossy items such as polished metal | Scenes with many separate objects |
| Furniture, vehicles, everyday items | Objects with fine text or patterns that matter | Flat drawings or screenshots viewed head-on |
| Illustrated characters with a clear pose | People in loose clothing or with hair blowing | Extreme close-ups that crop the object |
Photos of people and pets
For a figurine of a person, a full-body photo in a simple standing pose gives the best result: arms slightly away from the body so they do not merge with the torso, feet visible. Only upload photos of people who have agreed to it. For pets, get down to their level and shoot in even light; dark fur hides shape in shadow, so brighten the scene rather than the photo afterward.
Illustrations and concept art
Drawings work surprisingly well when they depict one object in a slight perspective view with clear shading. Purely flat, symmetrical front-view art produces a shallow result because there is no depth cue. Adding a rough shaded underside or a three-quarter pose helps. See also using extra angles when one view is not enough.
A ten-second test before you upload
Squint at your photo. Can you trace the outline of the subject with your finger without hesitation? Can you guess what the back looks like? If both answers are yes, the model probably can too. If not, change the photo before changing the settings; it is faster than waiting in the queue to find out.
Still not right? Read why your 3D result looks wrong for a symptom-by-symptom fix list.
Try it on your own image
Upload a photo and get a 3D model or a voxel build. Free, no sign-up.
Image to 3D Image to Voxel