A slicer is the least forgiving program in your pipeline. It doesn't care how the model looks in the viewport — it cares whether it can decide, for every point in space, whether that point is inside the object or outside it. This is a guide to the errors that stop it from deciding, how to find them before you waste a print, and which repair tools are still alive as of September 2026. Disclosure up front: we make Re:Form, a browser sculpting app, and it appears in the workflow sections — weigh those accordingly.
Why won't my 3D model print?
Most models fail in a slicer for one of five reasons: holes, non-manifold edges, flipped normals, self-intersecting geometry, or floating disconnected shells. All five confuse the same question — is this point inside the model or outside it? Thin walls and wrong scale cause most of the rest, and those slice fine and fail on the printer.
The distinction matters because the two groups have different fixes. Geometry errors are repaired with a repair tool. Thin walls and scale errors are design problems: the file is valid, and it will happily produce a fragile or wrongly-sized object.
What "watertight" and "manifold" mean
A watertight mesh has no holes. Pour water in and none escapes. Every edge in the surface is shared by exactly two faces, so the shell is closed.
A manifold mesh is a stricter idea: the surface behaves like real material everywhere. Every edge joins exactly two faces, and every vertex has a single continuous ring of faces around it. Two cubes touching at exactly one corner are watertight but non-manifold — at that vertex the surface pinches to zero thickness, and the slicer can't work out what to do with material that is infinitely thin.
Printers need both. What the slicer actually does is intersect your surface with a stack of horizontal planes and expect a closed loop at every level. Any place where the surface is open, doubled, or pinched breaks a loop.
Do sculpted models actually break?
Less often than CAD, kitbashed, or scanned models. A sculpt built by pushing one continuous surface stays closed by construction, and a uniform remesh rebuilds it as a single watertight shell. Sculpting's common failures are not holes but floating parts, walls too thin to print, and models arriving at the wrong size.
That's worth knowing before you go hunting for repairs you don't need. If your model came from a scanner, a boolean-heavy CAD assembly, a decimator, or a marketplace download of unknown parentage, assume geometry errors. If it came out of a sculpting app as one remeshed body, check the design problems first.
The five errors that stop a slice
| Error | What it looks like | Usual cause | Fix |
|---|---|---|---|
| Holes | Edges belonging to one face only | Deleted faces, failed boolean, incomplete scan | Fill holes, or remesh the whole surface |
| Non-manifold edges | An edge shared by three or more faces | Overlapping shells welded badly, mirrored halves merged at the seam | Remove interior faces, then re-weld |
| Flipped normals | Model renders inside-out or patchy; slicer prints a shell inverted | Mirroring without flipping, imported scan data | Recalculate normals outside |
| Self-intersections | Geometry passing through itself | Posed limbs pushed into the torso, overlapping merged parts | Union or remesh so the intersection becomes real surface |
| Floating shells | A separate island of geometry, sometimes tiny | Stray brush strokes, hidden layers, debris from booleans | Split by loose parts and delete the debris |
Flipped normals deserve one extra note: a slicer's auto-repair often "fixes" them by making a guess about which side is outside, and it guesses wrong on models with internal cavities. If a print comes out hollow where it should be solid, suspect normals.
How do you check a model before slicing it?
Load it into the slicer first — PrusaSlicer, Cura and Bambu Studio all flag broken geometry on import. For a real diagnosis, Blender's free 3D Print Toolbox extension is the standard: it checks manifoldness, flipped faces, wall thickness, overhang angle and intersecting geometry, and clicking a result selects the offending faces.
That last part is what makes the toolbox worth the detour. A slicer tells you a number. The toolbox shows you where, which is the difference between an informed fix and blind repair. It also ships a clean-up operator that fixes bad normals, fills holes, and removes empty edges and faces, plus a Hollow tool for giving a solid model consistent wall thickness — and from Blender 4.x there's a Make Manifold operation that runs the common repairs in one go.
If you don't want to install Blender for this, the slicer's own repair is the next stop.
What can the slicer fix on its own?
More than people expect. PrusaSlicer auto-repairs on import and adds a "Fix by Netfabb" command calling Microsoft's repair APIs — which Prusa's knowledge base documents as Windows-only, so Linux and macOS users need an external tool. Cura instead offers Mesh Fixes settings: "Union overlapping volumes" and "Remove all holes".
PrusaSlicer's automatic pass catches a surprising amount on its own, and Cura's two settings between them cover interfering solids and unwanted interior cavities.
Both approaches share a limitation. The slicer repairs the model for this slice; it doesn't hand you back a fixed file. If you plan to reprint, sell, or share the model, fix it at the source and re-export.
Which mesh repair tools still work in 2026?
Four routes are worth knowing, and one of the most-recommended is unmaintained. Autodesk stated in September 2021 that Meshmixer is no longer in development and will not be supported, and that hasn't changed as of September 2026 — the last release still runs on Windows and still appears on Prusa's recommended list, but nobody is fixing it.
- Blender + 3D Print Toolbox — free, actively maintained, and the only option here that also shows you what's wrong. Steepest learning curve, best outcome.
- Microsoft 3D Builder — Windows only. Prusa's knowledge base still lists it as an easy repair route; drag the file in, hit repair, export. Worth using if it's already on your machine.
- Meshmixer — auto-repair plus a "make solid" operation that voxelizes the model into a guaranteed-closed shell at the cost of surface resolution. Unmaintained since 2021, still effective, still widely used.
- Online repair services — Prusa's article points at Aspose and Formware's browser-based STL repair. Convenient, and worth a thought before you upload a model you intend to sell to a third-party server.
A fifth route is the one sculptors already have: remesh. Rebuilding the surface at a uniform resolution produces a single closed body and discards the topology that was broken, which is why "merge everything, then remesh" is the standard print-prep move rather than a repair step. It costs you the original topology, which matters for animation and not at all for printing. We wrote up the wider trade-off in remeshing versus retopology.
Thin walls fail more often than broken geometry
A valid mesh can still be unprintable. Formlabs' published design guidelines put minimum wall thickness at roughly 1 mm for FDM, 0.2 mm for SLA resin (supported or unsupported), and 0.6 mm vertical / 0.3 mm horizontal for SLS — with the note that FDM walls should be a multiple of your nozzle width, so 0.8 mm or 1.2 mm on a standard 0.4 mm nozzle rather than an awkward 1.0 mm.
| Process | Minimum wall (Formlabs guidance) |
|---|---|
| FDM | ~1 mm, in multiples of nozzle width |
| SLA / resin | ~0.2 mm supported and unsupported |
| SLS | 0.6 mm vertical, 0.3 mm horizontal |
Translate that into sculpting terms and it's a short list of usual suspects: cape and cloth edges, fingers, whiskers, antennae, weapon blades, ear rims. Every one of those reads beautifully on screen at any thickness, because pixels have no minimum. Measure them at print scale — a cloak edge that is 0.3 mm on a 32 mm miniature is not going to arrive.
The fix is boring and it works: thicken the offending parts before export. Inflating a fabric edge by a fraction of a millimeter is invisible in the final print and the difference between a blade and a snapped stub.
Does your model's scale survive the export?
Not by itself. STL stores triangles with no unit declaration, and OBJ is no better, so the slicer assumes — usually millimeters. That assumption is why models arrive 25.4 times too small. The 3MF format fixes it by declaring units in the file, millimeters by default, and PrusaSlicer, Cura and Bambu Studio all read it.
Until your sculpting app exports 3MF, the workaround is to decide scale before export rather than after. In Re:Form, the Grid section has a Unit mm slider that declares how many millimeters one world unit represents, and exports scale by it — set your sculpt to 40 mm per unit and a two-unit figure arrives in the slicer at 80 mm. Whatever tool you use, the habit that saves prints is the same: check the model's bounding-box dimensions in the slicer before you send it, every time. Our file format guide covers what else each format does and doesn't carry.
Floating parts, and the sculpting habit that creates them
Multi-part sculpts are the most common source of print failures that aren't geometry errors at all. You sculpt the body on one layer, horns on another, eyes on a third, hide a rough draft layer and forget it, then export — and the slicer receives four separate objects, two of which are floating in mid-air, plus a hidden blob you abandoned an hour ago.
The order that avoids this: split loose to find the debris, delete it, merge everything visible, remesh, then export. Splitting by loose parts first is the step people skip, and it's the one that surfaces the stray blob you can't see because it's inside the head. After merging, the parts still overlap rather than truly join — remeshing is what converts those intersections into a single continuous surface. Re:Form's print-prep documentation walks the same four moves in the app, and the broader browser workflow is in our sculpting for 3D printing guide.
A pre-export checklist that avoids repair entirely
- Split loose parts and delete the debris. Stray strokes and abandoned drafts, gone before they reach the slicer.
- Merge everything visible into one body. Including the layer you hid and forgot.
- Remesh. One continuous watertight surface; overlaps become real geometry.
- Measure the fragile bits at print scale. Anything under about 1 mm for FDM or 0.2 mm for resin gets thickened now, not after a failed print.
- Set real-world scale before export, then verify the bounding box in the slicer.
- Export STL or OBJ, and read the slicer's import warnings instead of clicking past them.
Six steps, maybe two minutes. That's cheaper than an eight-hour print that comes out hollow.
Where the browser fits
Re:Form does the print-prep half of this list natively: Split loose, Merge all, Remesh, a Unit mm scale control, and export. It's a browser 3D sculpting app on WebGPU — desktop Chrome, Edge, Firefox and Safari 26, nothing to install, a free sign-in and then you're sculpting. It is also alpha software (v0.7.2 as of September 2026), and it is not a mesh repair tool: it has no diagnostic that reports non-manifold edge counts the way Blender's toolbox does. For a model you sculpted, remeshing makes that diagnostic mostly moot. For a broken file someone sent you, Blender remains the right answer, and we'd rather say so.
The Free plan covers the whole printing pipeline for personal work — every brush, remeshing, roughly 12.5 million triangles, and OBJ export, which every major slicer reads. Pro at $9.99/month or $99/year adds STL, GLB and FBX, the 80-million-triangle budget, and the commercial license you need to sell prints. If you want to try the workflow on something small, open the app and sculpt a 32 mm figure; our tabletop miniatures guide has the scale numbers.
FAQ
Does remeshing guarantee a watertight model?
Remeshing rebuilds the surface as one continuous shell, which resolves overlaps, most holes, and inconsistent normals in a single step — it's the reason sculpted models rarely need a repair tool. It is not a universal repair: it discards your original topology, absorbs detail finer than the chosen resolution, and won't rescue a model whose parts were never meant to touch. Verify the result in your slicer before printing.
Do I need Pro to print from Re:Form?
No. The Free plan exports OBJ, and PrusaSlicer, Cura, Bambu Studio and Orca all open OBJ files without complaint. Pro ($9.99/month or $99/year) adds STL export, which is the slicer-native format and marginally more convenient, plus GLB and FBX, more than six times the triangle budget, and the commercial license required if you intend to sell prints or printed models.
How many triangles should a print-ready model have?
Fewer than you think. A printer resolves detail measured in tenths of a millimeter, so triangles finer than that add file size and slicing time and change nothing about the object that comes off the plate. A remesh resolution matched to your finest real detail beats maximum density every time. Our guide to how many polygons you actually need has the numbers per use case.