Fix an STL Online, Free — and What Free Actually Costs
Search for a free online STL repair tool and you get a long list, all saying roughly the same thing. They are not the same, and the differences are not in the feature list.
Drop your STL file here
or
Nothing is uploaded — the repair runs on your own machine.
Three questions the word free doesn't answer
Every tool in this category is free, requires no signup, and claims to fix holes and non-manifold geometry. Those claims are mostly true and mostly identical, which makes them useless for choosing. These three are what actually differ.
1. Where does your file go?
There are two architectures behind these tools, and the page rarely says which one you are using.
Server-side: your model uploads, gets processed on someone's machine, and comes back. This is why some tools can promise that a slow computer doesn't matter — the work isn't happening on your computer. The trade is that your geometry left your machine. For a downloaded trinket that is irrelevant. For a client part, an unreleased product, or anything under NDA, it is the whole question.
Client-side: the repair engine is compiled to WebAssembly and runs inside your browser tab. Nothing is transmitted. The trade is that your own CPU and your browser's memory ceiling become the limit.
A quick way to tell which you are on: load the page, disconnect from the internet, then try to repair a file. Client-side tools keep working. It takes ten seconds and settles the question better than any privacy paragraph.
2. What size actually works?
Stated file-size limits are often the size of file the upload form accepts, not the size it can meaningfully finish. A 100 MB limit sounds generous, but a 100 MB STL is roughly two million triangles, and mesh repair does not scale linearly with triangle count — the work grows faster than the file does.
In practice that means a tool can accept your file, spin for several minutes, and either finish long after you have given up or run out of memory. The limit that matters is not what the form accepts; it is what finishes in a time you are willing to wait.
This is a physical constraint, not a quality difference between tools. Anything running in a browser tab shares the same memory ceiling. What varies is whether you are told before you commit or after you have waited.
3. Are you told what changed?
Most of these tools return a file and nothing else. You cannot tell whether it sealed four holes or rebuilt half your model, whether it succeeded, or whether it quietly skipped a check because the mesh was large.
This matters more than it sounds, because the failure mode is silent. A repair that didn't fully work still hands you a file. You find out when the slicer refuses it again, or worse, when a print fails nine hours in with a hollow section.
Our answers, including the limits
This tool is client-side: the engine runs in your browser and your file is never uploaded. You can verify that with the disconnect test above.
On size, it reads your file's triangle count before doing any work and tells you the estimated time upfront — a few seconds for a typical model, around half a minute for a very dense one. Past roughly 2.5 million triangles it declines rather than letting you wait for a failure, and points you at desktop software instead.
On reporting, you get the repair log in plain language: holes sealed, faces re-oriented, junk triangles removed, whether the result is watertight, how many separate shells it contains. And when a check is skipped for speed on a large mesh, the report says the check was not run rather than showing a clean tick.
What it will not do: reconstruct geometry that was never in the file. Holes are closed with flat patches, so a large missing region comes back bridged rather than rebuilt. It also does not sculpt, hollow, or split models — it does one job.
Frequently asked questions
Is a free online STL repair tool safe to use with confidential models?
It depends entirely on whether the tool processes your file on a server or in your browser, and the page does not always say. The practical test: load the page, disconnect from the internet, and try repairing a file. If it still works, nothing is being uploaded. If it fails, your file was going somewhere.
Why do tools advertise 100 MB limits if that size doesn't work well?
The stated limit is usually what the upload form accepts, not what finishes in reasonable time. A 100 MB STL is around two million triangles, and repair effort grows faster than file size. The number that matters is how long it takes, which is why this tool shows you an estimate before starting rather than a maximum file size.
Is free worse than paid repair software?
For the specific job of making a broken mesh printable, no — the underlying algorithms are well understood and available in open-source libraries. Paid tools like Netfabb earn their price on build preparation, packing and production workflows, not on basic mesh repair.
How do I know the repair actually worked?
Re-import the file into your slicer. If the warning is gone and the preview shows continuous walls with infill inside them, it worked. That test is more meaningful than any number a repair tool reports, including ours.