Decimation reduces the amount of geometry. Retopology reorganizes the geometry's connectivity. Both can change polygon counts, but they solve different problems. Choosing one by the file's triangle count alone often leads to disappointing results.
If your goal is to make a static web model faster, a measured reduction may be enough. If you need cleaner edge flow for editing or deformation, a new topology may be more useful—but it takes more review, and automatic results are not guaranteed to be acceptable.
What decimation is good at
Decimation removes or collapses parts of a mesh to use fewer polygons while trying to preserve the surface's visual shape. It is usually chosen for simplified LOD meshes, background props, scan cleanup stages and other situations where rendering or storage cost matters more than deliberate edge layout.
The main tradeoff is detail. Thin pieces, hard edges, silhouette corners, UV islands and material boundaries can degrade after aggressive reduction. A smaller mesh is not automatically faster in every scene; large textures, many materials and expensive shaders still matter.
Decimation does not promise edge loops that deform naturally at a character's elbow or knee.
What retopology is good at
Retopology rebuilds the mesh with a different arrangement of faces and edges. A quad-dominant topology can be easier to edit in a modeling application and may create a cleaner starting point for later operations. For characters, deliberate loops around joints are particularly important; a generic automatic solution may not place them where a rigger needs them.
Auto-retopology engines have to infer how to redistribute faces from existing geometry. Complex objects with many separate parts, open edges, thin surfaces, or intersecting components are harder than a clean single shell. A result can look superficially similar but still be unsuitable for UV work, animation, or further authoring.
Retopology can lower or raise face counts depending on the target topology and settings. Its defining property is a new mesh structure, not simply 'fewer triangles.'
A practical comparison
- Main objective: Decimation: Reduce mesh density; Retopology: Rebuild topology / edge layout.
- Typical input: Decimation: Dense mesh; Retopology: Mesh needing different structure.
- Expected output: Decimation: Lower-density geometry; Retopology: Reorganized geometry, often quad-dominant.
- Deformation-ready?: Decimation: Not by itself; Retopology: Only after deliberate quality checks.
- Main visual risk: Decimation: Loss of detail and silhouette; Retopology: Shape drift and changed surface features.
- UVs/materials: Decimation: May require checks or rebaking; Retopology: Often requires substantial UV/material transfer review.
Neither method is a replacement for validating the actual output in the context where you will use it.
Why a clean input matters
Before processing, inspect whether your mesh contains duplicate or degenerate faces, disconnected shells, unexpected open boundaries, or very thin intersecting components. Some are genuine defects; others are legitimate modeling choices. A vehicle with separate wheels and body panels should not be turned into one melted shell merely to satisfy an algorithm.
Record the original silhouette and expected dimensions. Keep a backup. After processing, inspect key details and any material or UV changes alongside the new topology.
The 3D Model Analyzer helps surface geometry and topology information before making a decision. Review the existing web model optimization checklist when your real goal is runtime performance.
Where 3DLabTools Auto Retopology fits
3DLabTools Auto Retopology is currently presented as an Alpha/Experimental workspace for exploring alternative topology on local static geometry. Treat candidate meshes as reviewable experiments, not guaranteed production assets. It accepts GLB, embedded glTF and FBX input, then performs its own geometry admission checks. Source textures and material assignments are not transferred to the neutral output, and rigging, morph targets and animations are not preserved. The current shell strategy can preserve small or thin geometry rather than rebuilding every part.
For a character with animation requirements, choose an appropriate rigging-aware authoring process and validate deformation separately. Do not assume experimental static-mesh retopology will preserve animation, skin weights or every UV detail.
An illustrative decision walkthrough
Imagine a textured static prop whose silhouette is already good but whose rendering budget is too high. First separate triangle cost from material count and texture memory in Analyzer; its draw-call and texture-memory figures are estimates. If triangles are the relevant cost, try the separate LOD Generator on a static GLB and compare each level. If face layout itself is the problem, explore Retopology Alpha on a copy, knowing that its neutral output is not a textured replacement. This is a diagnostic example, not a measured speed or compression claim.
Which approach should you choose?
Choose decimation when you primarily need fewer polygons, understand the acceptable visual loss and can validate runtime improvement. Investigate retopology when face layout itself limits editing or later processing. Consider leaving the mesh unchanged when the measured problem is actually texture memory, shader cost or draw calls.
Next step: Inspect the asset with the Model Analyzer, then explore Auto Retopology Alpha on a disposable copy if a topology rebuild is the right experiment.
Verification note: reviewed against the current static-mesh tool boundaries and Alpha output limitations on October 20, 2026.
Explore product documentation or open the Auto Retopology Alpha (Experimental).
