A 3D scan can look beautiful and still be too heavy for a browser. This beginner tutorial explains mesh topology, why scans often contain too many vertices, and how decimation can make a model faster without making it look worse.
Mesh topology is simply the structure that makes up a 3D model. You do not need to be a 3D artist to understand the basics.
Vertices are points in 3D space. A scanned model can contain millions of them, even where the shape is almost flat.
Edges connect vertices. Together they form the wire-like structure you see when a model is displayed as a mesh.
Faces are the small surface pieces between edges. In many scanned models, these faces are triangles.
Photogrammetry and 3D scanning software reconstructs surfaces from photos or sensor data. The software does not always know which parts of the scene are important to a human viewer. Because of this, it can spend thousands of tiny triangles on a wall, floor, table, or other flat area where only a few triangles would describe the same visible plane.
This does not mean the scan is bad. It means the raw scan is usually a production model, not a web delivery model. Before publishing, it is normal to create a lighter copy that keeps the visible shape but removes unnecessary geometry.
A desktop 3D application and a web browser are not the same environment. Web scenes need to load quickly and stay responsive.
More vertices usually mean larger files. Visitors may wait too long before they can even see the scene.
The browser must keep geometry, textures, and viewer data in memory. Heavy meshes can push devices over their limits.
Phones and tablets have less graphics power than workstations. Very dense models can stutter, overheat, or fail to open.
Orbiting, zooming, clicking points of interest, and moving through a scene should feel smooth for the viewer.
For general web publishing, aim for around 1 million vertices. This is not a strict rule, but it is a useful starting point for models that should work well across desktop and mobile devices.
Decimation reduces the number of vertices and faces in a mesh while trying to preserve the overall shape.
Beginners often worry that reducing geometry automatically means reducing quality. In practice, a scanned model usually contains a lot of geometry that no viewer can notice. If a flat wall uses 50,000 tiny triangles, it can often be reduced dramatically and still look like the same wall.
The goal is not to make the model as small as possible. The goal is to remove geometry that does not improve the visible result.
If your scan has 10 million vertices, try a first optimized version around 2 million, then move closer to 1 million if it still looks good.
Check the silhouette, thin details, carved edges, and important surface features. Flat areas can usually lose much more geometry.
Upload the optimized copy and open it on a normal laptop and phone. The best target is the one that looks good and feels smooth.
Different software uses slightly different words. Blender calls it Decimate, Metashape calls it Decimate Mesh, and RealityScan calls it Simplify.
Tip: If the model is mostly flat architectural surfaces, test Planar mode too. It can remove extra faces from flat areas.
Tip: If the dialog asks for faces rather than vertices, do not worry about an exact conversion. Aim for the recommended range, export, then check the model statistics and visual result.
Tip: RealityScan creates a new simplified model instead of overwriting the original, which is useful when comparing versions.
Use this quick checklist before publishing a scanned model in Kroscloud.
Some projects really do need a very detailed model. For example, an internal desktop-only review, a specialist research workflow, or a controlled presentation may benefit from higher complexity. Kroscloud does not block those workflows with a hard vertex limit. For most public web scenes, however, an optimized model gives visitors a much better experience.
Mesh topology is the structure of a 3D model: vertices, edges, and faces. A clean model uses enough topology to describe the shape, but not millions of unnecessary points on simple surfaces.
Scanning and photogrammetry software reconstructs the surface from many measurements or photos. It often creates dense geometry everywhere, including flat areas where the extra triangles do not improve the visible result.
Not necessarily. Good decimation removes geometry that is hard or impossible to see. If you reduce the model gradually and check the result, the web version can look very close to the original.
No. Around 1 million vertices is a practical starting point for general web viewing, not a strict rule. Smaller models may work with far fewer vertices. Specialist desktop-only models may use more.
Kroscloud does not apply a hard mesh complexity limit because some professional workflows need very detailed models. The recommendation to optimize is about performance and visitor experience, especially on mobile devices.
For very dense scans, start with geometry decimation. After that, check texture size and format. A balanced model uses reasonable geometry and textures that are detailed enough but not unnecessarily huge.
Create a lightweight copy of your scan, upload it, and test how it feels as an interactive browser scene.