Corpo Humano Por Fora

Getting a clean external scan of the human body is harder than people think

I spent about three years working with structured-light scanners and photogrammetry rigs before I stopped throwing money at equipment I didn't need. The goal is usually to capture the human form externally — skin surface, posture, clothing folds, everything visible from the outside — without the noise that comes along with it. It sounds straightforward until you try it.

corpo humano por fora: the actual workflow

Start with a turntable setup if you're doing this solo. A motorized platform that rotates at a constant speed saves you from chasing the subject around. Pair that with two cameras on a rig at roughly shoulder height, aimed at the center axis. Natural light out a window works, but consistent continuous lighting is cheaper than dealing with clouds moving across the sky over a forty-minute capture session. The actual capture process runs like this: position the subject standing still in a neutral stance with arms slightly away from the torso. Take overlapping photos every thirty degrees of rotation. You need at least sixty images, but seventy-five gives you breathing room for post-processing. Each photo should overlap the previous one by about sixty percent. That overlap is non-negotiable — drop below fifty and your reconstruction software will start dropping sections of the model.

For structured-light approaches, project a known pattern onto the subject and capture the distortion. This works faster than photogrammetry but demands a darkened room and careful calibration. I switched from photogrammetry to structured light for client work because the turnaround time dropped from about ninety minutes per scan to roughly twenty minutes, though the equipment cost went up significantly.

the edge case that almost broke me

One specific problem I ran into was scanning subjects with dark hair against dark clothing. The structure-from-motion algorithms couldn't find enough feature points in those regions, so the reconstruction would produce complete voids — gaps where the hair met the shoulders, or where a black shirt blended into the background. No amount of retaking photos fixed it because the issue isn't lighting, it's textureless surfaces. The workaround I settled on was applying a light dusting of talcum powder or using a commercial scanning spray like Artec Scanning Spray. A thin, even coat creates micro-texture that the cameras can latch onto without looking obvious in the final render. The spray dries in about thirty seconds and washes off with soap and water. Talcum powder is cheaper but harder to remove completely from fine fabrics and tends to shift during the scan, which introduces noise into the mesh. I use the spray now as my default and only reach for powder when budget is tight.

what most beginners miss about external body scanning

The first thing people get wrong is assuming the scan itself produces a usable model. It doesn't. Raw point clouds from a body scan are dense but full of artifacts — floating particles, stray noise from sweat or skin oils, and the characteristic "mushroom" artifacts around joints where the scanner loses tracking during subtle movements. Cleaning these takes more time than the capture. The second common mistake is underestimating how much the subject moves. Even breathing creates enough displacement to corrupt a high-resolution scan. I've seen people spend two hours scanning someone who was visibly trembling from nervousness. The result looked like a blurred photograph translated into geometry. The fix is simple but unpleasant: ask the subject to hold their breath during each capture burst, or better yet, use a continuous scan mode that averages thousands of frames over several seconds. My preferred rig captures at about twelve frames per second and the software averages them down to a single clean mesh.

A third thing nobody warns you about is the handling of extremities. Fingers, toes, and the space between the thighs tend to collapse or merge together in reconstruction. The scanner simply can't resolve geometry at that scale without physically separating those regions, which is obviously not acceptable in most professional contexts. I got around this by scanning the subject twice — once in a standard stance and once with strategic padding between fingers and toes using medical-grade foam spacers. The two scans merge cleanly in the processing stage because the underlying geometry matches.

software and practical recommendations

For photogrammetry pipelines, RealityCapture handles body scans better than most alternatives. It processes a full set of one hundred images in about eight minutes on a machine with a decent GPU, and the mesh quality is generally clean enough to skip heavy manual repair. Meshroom is free but slower and produces noisier results on human subjects specifically. If you're doing structured-light scanning,Artec Studio remains the industry standard despite the licensing cost. The automatic alignment and fusion pipeline is hard to beat for speed, and the built-in skin smoothing tools save you from manually fixing half the artifacts that come through raw.

For export, aim for OBJ or GLB formats. PLY works for intermediate processing but most downstream applications — game engines, AR platforms, animation pipelines — prefer one of the other two. GLB is the better choice if the model needs to be viewed in a browser or lightweight viewer. It keeps textures and geometry bundled in a single file.

corpo humano por fora: knowing when to walk away

There are situations where external body scanning simply won't give you what you need. Subjects wearing loose or flowing garments produce mesh that looks correct geometrically but has no meaningful surface detail — the fabric folds become the dominant feature and the underlying body shape gets buried. This isn't a processing problem, it's a physical limitation. You can't reconstruct skin texture through layers of cloth. People with significant body hair also present challenges for structured-light systems. The light pattern scatters across individual strands rather than reflecting cleanly off the skin surface, creating a noisy, inflated mesh that looks like the subject gained ten pounds during the scan. Photogrammetry handles this better because it relies on color and contrast rather than light pattern distortion, but resolution drops noticeably.

If your end goal is animation rigging or motion capture, consider whether a full external scan is actually necessary. A single neutral-pose scan provides visual reference but nothing about how the body moves. Combining a surface scan with a skeleton rig built from proportional measurements is often more useful than a pristine mesh that sits idle. I learned this the hard way after delivering what I thought was a premium scan product and having the client return it because they needed pose data, not just geometry. The equipment list for a functional setup runs roughly four to six thousand dollars depending on whether you buy new or refurbished. Cameras, lighting, turntable, and software licenses are the main cost drivers. Anything advertised as a complete solution under two thousand dollars will compromise either resolution or speed in ways that matter once you start working with real clients. Budget accordingly.