If I said I was finished, I’d be lion.

After a lovely month in the UK, and some time in Ireland and the Netherlands, it’s been slow moving on the sub-d Unity project.

On the plus side, I have edge weighting “working” now!

So that’s my old GoLion/Voltron model from Modo, but all I brought into Unity was the low poly mesh (on the left), and generated the two right models in C# in Unity using Catmull-Clark sub-division with edge weighting.

You can see that the higher LODs still keep semi-sharp edges, rather than just rounding out completely, thanks to the edge weighting.

Some of the things left to play around with:

Edge loop reduction on high lods (to make vert counts reasonable).

Support for subdividing all the other vertex data (normals, vertex colours, UVs, arbitrary additional vertex data)

Vertex positions are about 1% different to Modos. I’m presuming this is because I’m not using Chakin’s curve algorithm to alter edge weights per division (it’s the only part of the Pixar document I ignored, so it’s a fair chance that is it)

Create “mid” lods, where every Edge point and Control point has enough information to transition back to the appropriate position on the control surface, so that I can blend smoothly from one lod to another.

***

I might not go any further with this, but it has been a fun exercise.
This sort of approach *might* be useful for non interactive structural meshes in levels.
And probably not for much else (unless you’re living in the past, and targeting DX9 hardware, maaaaan) 😉

Honestly, OpenSubDiv is the way to go, I can’t wait for all the software I own to have it 🙂