Fixing "Non-Manifold" Errors from Your Slicer
"Model is not manifold." "This object is not watertight." "Auto-repaired mesh — results may vary." Every slicer phrases it differently, and none of them tells you what is actually wrong with the file or what the auto-repair just did to it. This guide translates the error, tells you when you can safely ignore it, and gives the two-minute path from warning to a mesh that slices without complaint.
What the Error Means
A slicer computes toolpaths by asking, at every layer, which regions are inside the part. That question only has an answer when the mesh is manifold: every edge shared by exactly two triangles, so the surface cleanly separates inside from outside. Two defects break it — boundary edges (an edge with only one triangle: a hole) and non-manifold edges (three or more triangles meeting at one edge: an ambiguous junction, usually an internal wall left over from a boolean operation). The slicer flags the file because for some region it genuinely cannot tell what is solid.
| Slicer message | Usual meaning |
|---|---|
| Cura: grey/striped areas in preview | Regions whose inside could not be determined; often flipped normals |
| PrusaSlicer: "object has errors" (triangle icon) | Counts non-manifold edges and holes; offers Netfabb repair on Windows |
| Bambu Studio: "Auto-repaired" | Repair already applied silently; check what it changed |
| Any slicer: part slices but walls are missing | Flipped normals or internal faces confusing the inside test |
Can You Ignore It?
Sometimes, honestly. Modern slicers auto-repair on import, and for a small hole or a stray internal face their guess is usually right. The gamble is that you only find out at layer 143 when a wall is missing. Ignore the warning when the part is decorative and the preview looks right at several heights. Do not ignore it when the part is functional, when the preview shows any anomaly, or when the same file will be printed more than once — two minutes of checking beats re-printing.
The Two-Minute Fix
1. Upload the file to the checker in check mode. The report counts the actual defects: boundary edges, non-manifold edges, flipped winding, duplicate faces, shells. Now you know what the slicer was complaining about, and the 3D preview shows where.
2. Run repair mode at the Standard level. Welding, winding fixes, hole filling, and reconstruction clear the common defects; the report prints before/after numbers including dimensions and volume, so you can verify nothing drifted.
3. Re-slice the output. The warning should be gone; if residual defects remain the report says so explicitly rather than pretending, and the aggressive level (which may move geometry slightly, and says when it did) is the next step for hairline cracks.
Paste the file your slicer rejected and read the actual defect count — check mode never modifies it.
Check & Repair the STLStopping It at the Source
The same errors keep coming back when the source keeps producing them. Boolean unions that leave internal walls: merge solids properly in CAD instead of stacking overlapping bodies. Meshes combined in a mesh editor without a union: same disease, mesh flavor. Downloaded models: check before printing, the upload was probably never validated. And exports at extremely loose tolerance can crack seams at shared edges — the export tolerance guide covers picking a sane value.
Frequently Asked Questions
Why does my slicer say non-manifold when the model looks fine?
Rendering and slicing ask different questions. A viewer just draws triangles, and a mesh with internal walls, tiny holes, or flipped patches draws perfectly. The slicer has to decide inside versus outside at every layer, and it is the geometry underneath — edges shared by the wrong number of triangles — that breaks, invisibly to the eye.
Is it safe to let the slicer auto-repair?
For throwaway prints, usually. The problem is silence: the slicer does not report what it changed, so a filled hole that should have been a slot, or a removed wall that should have stayed, only shows up in plastic. A repair with a before/after report gives you the same fix plus the evidence.
What causes non-manifold geometry in the first place?
Mostly booleans that leave internal faces, meshes merged without a union, damaged downloads, and scan output. In CAD-sourced files it is almost always the boolean history; in sculpted or scanned files it is the capture process itself.
My file has thousands of non-manifold edges. Is it hopeless?
Usually not — large counts typically mean one systematic cause (every seam cracked by a loose export, or a whole internal shell), and one reconstruction pass clears them all. The unfixable case is an open surface with no thickness, and a good checker tells you that directly instead of trying.
This guide covers one problem. For the full set of 3D converters, checkers, and calculators, visit the hub.
Browse 3D Tools