I'm thinking about employing some basic LOD in my renderer, by having low detail versions of my models for drawing in the distance. I just need a bit of advice on the following: - Who is responsible for creating the lower-detail versions of models, the programmer (me) or the artist? If it's my resposibility (doh!) are there any common approaches to this problem I should know about? - When using vertex arrays / buffers, should each LOD have its own vertex and index arrays, or would it suffice to use a common vertex array and give each LOD its own unique index array? Any general advice about beginning to implement LOD is also very welcome! Thanks [smile]

The artist creates the LODs and the programmer writes the code that switches between them.

The programmer could write software that processes a model and generates the LODs, but the results will be poor for most models.

Thanks! That's a bit of a relief. Anyone able to inform me about the second question?

Quote:Original post by The Rug
Thanks! That's a bit of a relief. Anyone able to inform me about the second question?

Since they are separate models, it is probably better to use separate vertex and index buffers.

If you're really keen :), there are some advanced techniques for calculating salency. This is a measure of the overall importance of each vertex in a mesh, *i think* :). This can then be used to gracefully produced LOD's automatically. If you're interested check out this years SIGGRAPH papers, but then again it'll take a lot of work and much maths... Much beyond my own capabilities ATM :(

On the other hand, you could use a library that has progressive meshes or automatic LOD builders, like D3DX or The GNU Triangulated Surface Library. These are pretty good at continuous LOD meshes and generating discreet, ie progressive meshes vs a single mesh for each LOD level respectively.

Hope it helps,
Andrew