Image to model
Realitycapture photogrammetry turns overlapping photographs into measured 3D data. This guide explains the capture logic, processing stages, useful outputs, and trade-offs before you begin.
Explore the workflow
Photogrammetry works best when the scene, camera coverage, and desired output are planned together. These related guides help you choose a practical starting point.
Mechanism checks
Photogrammetry is powerful, but it does not recover information that the camera never captured. Plan around these edges instead of treating every image set as equally usable.
Glass, polished metal, and clear plastic may not provide stable visual features for matching between photographs.
Workaround
Use temporary matte treatment where appropriate, change the lighting, or capture a separate reference geometry pass.
A reconstruction cannot reliably model the back, underside, or interior of an object when those areas are absent from the images.
Workaround
Add camera positions that expose concealed surfaces and keep coverage consistent around the subject.
Plain walls, smooth painted surfaces, and repetitive patterns can give the alignment process too few unique points to track.
Workaround
Improve image overlap, add non-permanent visual references, or combine the capture with measured control points.
Photographs can produce a convincing shape while leaving real-world scale ambiguous unless a known measurement is included.
Workaround
Add a scale bar, surveyed markers, or another dimension that can be applied during inspection and export.
Core sequence
A dependable result comes from treating capture, alignment, and model generation as connected stages rather than one automatic button.
Photograph the subject from a steady path with consistent exposure and enough overlap for neighboring images to share recognizable features.
The software compares shared features, estimates camera positions, and builds a sparse representation. Remove blurry or poorly connected images before continuing.
Generate denser geometry, apply texture when useful, then inspect holes, scale, edges, and unwanted background before exporting.
Visible transformation
The visual change is not magic: image overlap supplies evidence, alignment establishes camera relationships, and reconstruction turns that evidence into geometry.
Photo set
Reconstructed mesh
Capability split
The same project can be judged by what the image set provides and by what a manual modeling workflow can control. This comparison clarifies where image-based reconstruction is strongest.
Starting material
Image-based reconstruction
Overlapping photographs of the real subject
Manual 3D modeling
Reference drawings, measurements, or concept guidance
Best fit
Image-based reconstruction
Existing objects, spaces, terrain, and visible surfaces
Manual 3D modeling
Designed assets, clean forms, and controlled proportions
Surface detail
Image-based reconstruction
Can preserve captured texture and wear
Manual 3D modeling
Must be modeled or painted deliberately
Hidden geometry
Image-based reconstruction
Limited by what the cameras can see
Manual 3D modeling
Can be created even without photographic coverage
Scale control
Image-based reconstruction
Needs a known reference or measured control
Manual 3D modeling
Can be set directly during modeling
Repeatability
Image-based reconstruction
Depends on consistent capture conditions
Manual 3D modeling
Depends on a repeatable modeling specification
Production speed
Image-based reconstruction
Efficient when coverage is clear and the subject already exists
Manual 3D modeling
Efficient for simple, designed, or highly edited forms
Practical applications
Different teams value different parts of the reconstruction process. The right use case is one where photographic evidence is more useful than starting from an empty scene.
A museum team needs a digital record of a sculpture with visible wear and irregular surface detail.
The image set preserves appearance for inspection, documentation, and later presentation.
reality capture freeA drone operator captures overlapping views of a site, stockpile, or structure from planned flight lines.
The resulting spatial data gives the team a visual base for review, measurement, and communication.
reality capture drone mappingAn artist needs real-world rocks, buildings, or surface materials as a starting point for a scene.
Captured geometry and texture reduce manual reference gathering before cleanup and creative editing.
realitycapture tutorialA project lead must decide whether image-based reconstruction or manual modeling better fits the asset.
The comparison makes coverage, scale, hidden geometry, and editing requirements explicit.
meshroom vs reality capturePipeline checkpoints
These checkpoints describe the main evidence chain from capture to inspected asset. They are useful for planning reviews and finding problems before export.
Common questions
A few practical questions come up before the first image set is processed.
It is an image-based method for reconstructing 3D geometry from overlapping photographs. The software identifies shared visual features, estimates camera positions, and uses that evidence to build a point cloud, mesh, or textured model.
There is no single number that works for every subject. The useful goal is consistent overlap from enough viewpoints to cover every visible surface, with extra images added around edges, complex shapes, and areas that are easy to hide.
It can create a detailed and useful model when the images are sharp, well covered, and supported by a known scale reference. Accuracy falls when surfaces are reflective, transparent, textureless, moving, or missing from the camera views.
Use stable focus, even exposure, clear visual features, and a steady path around the subject. Avoid motion between shots, large gaps in coverage, and major lighting changes that make the same feature look unrelated across images.
No. The process can only infer geometry from available evidence, so an unseen underside or interior may be missing or unreliable. Capture additional angles or combine the result with manual modeling when hidden areas matter.