Geometry Nodes - Merging Mesh Borders with Uneven Vertex Counts?

[Nicholas_Hodgson]

Hi everyone,

I’m currently developing a tool using Geometry Nodes that’s designed to fuse or boolean decals onto a base mesh. I’ve hit a roadblock when trying to cleanly merge two separate pieces of geometry whose border edges have mismatched vertex counts. My goal is to preserve the original topology of both meshes as much as possible.

What I’ve Tried So Far:

Merge by Distance
I considered selecting both border edge loops and increasing the merge threshold — but this tends to collapse vertices unpredictably, resulting in stretched or degenerate triangles. Not ideal for clean topology.

Extrusion + Boolean
Another idea: extrude the outer border slightly to create a thin “collar,” then use it to boolean-cut — theoretically matching vertex density before merging. Unfortunately, this approach proved unreliable, especially on curved surfaces, where thin booleans tend to fail or create artifacts.

Subdivision + Vertex Snapping
I also explored subdividing the inner border heavily, then sampling the nearest vertices from the outer loop and snapping them into place. However, I haven’t found a reliable way within Geometry Nodes to subdivide only the border edge loop with precise control — let alone sample and reposition vertices based on proximity.

I’d be incredibly grateful for any suggestions, alternative workflows, or clever node setups that could help achieve a clean, topology-preserving merge between mismatched borders.

Thanks so much for your time and insight!

[Okidoki]

Just a very quick idea:
If you consider the completely “inner faces” to be deleted and some “on the edge” to be triangles like so:

…then there might be a “better match” for the olmost fitting ones and the triangle edge to be divided to match ?? …if you know what i’m saying… ?

[Nicholas_Hodgson]

Thanks for the response.
I get what you are saying but this system needs to work for any given topology. That solution would only work for this example.

[Okidoki]

Well… of course it was a only a quick idea maybe to categorize into “full adaptable outer” faces and “half usable” faces…
…also in general the given problem is to find any fitting vertices which does not mean that this might also give fittiing faces. If the the vertex counf does not fit then there is are proberly some need of using triangles.

On the other hand you wrote nothing about how you are trying to select anything at all… in fact i do not know how using any proximity values could lead to some pairing at all. There are the almost perfect fits… but then there also those who might be mapped to the one or the other neighbour and so changing the original topology.

For the example of round to rectangular or some more stepped diamond shape usually one tries to blend those forms… but the computer “does not recognise” any symmetry with some bz far more math…

So the question would also be: which…

@Nicholas_Hodgson

boolean decals onto a base mesh

…do you plan to use for this ??

Any ellipsoide or oval decal will always not match some more grid like base…

You might be able to select some rectangular part of the original mesh, circularize and somehow shrinkwrap it back to the basemesh. Here for some inner adaption…

saw this after making this: maybe only halfway to the other form so that the triangulated edge bridge is not only round

[Nicholas_Hodgson]

Sorry if I wasn’t clear. Everything relating to the insertion/boolean has been solved. The only problem I have is getting the vertex count around the decal border edge matching closely with the border of the hole that has been boolean/cut out.

I have both selections for the interior border edge and also the boolean border edge so using them as selections isn’t an issue.

The current workflow is to take the decal, use the border edge to fill a face and use that to boolean into a surface to create a hole. This means the hole will always match very closely to the decal.

My ideal solution would be to subdivide the border edge of the decal and then match the positions to the nearest index of the boolean border edge but I’m unsure if this kind of subdivision possible in geometry nodes (like the image below)

Edit: I think I found a node group that might do the trick here. Will give it a try tonight:

Subdividing edges in geometry nodes

Hi! I thought this was an interesting problem, so I tried my hands at building a “Subdivide Edges” node group, that let’s you specify the edges to cut and number of cuts per edge: [subdivide-edges-node] I build the node group in 4.2, but I don’t think I used any of the 4.2 specific features. Here’s a .blend-file, if you want to give it a try: Subdivide Edges Node.blend (1.1 MB) It works as follows: It breaks the mesh apart into its faces Then it replaces each edge of the faces with a cur…

With you last suggestion, this wont work as the decals wont always be elliptical.

[Hadriscus]

You’re looking for the holy grail, that kind of thing only exists in MeshFusion afaik. I don’t think you can get there without a good deal of research. Just trying to set expectations. Let’s take it one step back : why is the output of boolean not suitable ?

[Nicholas_Hodgson]

Its very possible… I just need to upres the border edges of the decal to match the boolean approximately and I can fuse them without artifacts.

I cant just simply boolean because the system needs to use open edges for the decal to fuse into the target geometry. Here’s an example of a decal to give you an idea of what I am talking about. As you can see it has an open edge:

[Okidoki]

Ahhh… so you do not want to bridge between the objects but somehow uses the decal margin as a cutout and want to subdivide the cutted margin to join this while also shrinkwrapping some planar base (with the real decal object parented to this) of the cutout to the other object.

I thought more of planar decals… and not of some well known hardsurface decal object addons…

[Nicholas_Hodgson]

No worries, my fault for not being clear. Appreciate all the responses regardless

I ended up solving the issue by using the nodegroup provided in this thread to subdivide the decal border edges a few times. I then captured the position of the boolean border edges and snapped the decal border verts to them. Seems like a very reliable method.