Modelling Question, easier to increase Polys or decrease it?

[Manakel]

Hello All, from a modeling and game point of view.
Is it easier to start with low poly rigged models and increase details of the model later ?
or
Is it easier to start with High poly models and decrease details of the model later ?

I have a 500polys character set or a 20 000 polys character set
in both case, they are already textured.

[birukoff]

From my small experience with modelers, it is better to have a high-poly model, and then decrease its poly-count. The reason is that high-poly model has more details, and when you produce a low-poly version, you simply remove these details. While if you start with low-poly model, it doesn’t have all these details at all, so you have to create them. It is more difficult.

[Crimity]

High to low has always been the easiest for me…

If you’re talking about for LOD, one big reason starting with the lowest lod might be bad is that your mesh is going to change in 3d space as you cut up the polys and move the new verts around, but the uvs will be pretty much the same if you want the lods to use the same texture space… for lower LODs you can get away with weird texture coordinates… but with the highest level, uv stretching will be very obvious.

If you actually meant for normal mapping… I always start with the the high poly version because I usually end up trying something out that might need some extra polys on low poly mesh… if I had a low poly mesh already done and uv mapped, I’d have to make changes to both for any major tweak I did…

Also… while there are programs/plugins that reduce detail (while usually not that well)… there isn’t really anything that is going to add detail the way you want automatically…

[rdb]

It’s best to keep either the high-poly one or both, and when it really becomes a problem (not likely, unless you’re using big pixel shaders) you can always use a lower resolution model.

Though, polycounts don’t matter that much, when talking about performance.

Inexperienced game programmers, when they see that their programs are running slow, often react by trying to reduce the number of polygons. This almost never makes any difference whatsoever. Back in the mid-90’s, reducing polygon counts was a reasonable strategy to make games faster. That strategy just doesn’t work any more. The reason for this is that video card manufacturers have been increasing the polygon limits of their video cards by leaps and bounds. In fact, they’ve raised the polygon limit so high that you almost never hit that limit any more: you usually hit some other limit first.

That’s the key to modern performance tuning: knowing all the limits of the machine other than the polygon limit, and figuring out which of those limitations you’ve run into. In other words, the key to modern performance tuning is diagnosis.

Taken from:
panda3d.org/manual/index.php/Performance_Tuning

[T_Rex]

That statement really needs to be kept in context, depending on just how high-poly the models are, and if you care whether your game is playable on older hardware.

While we’re on the subject, does anyone know what the typical poly count of a character model is on an average Xbox, PS2, or Gamecube game? David?

[TheClaus]

I found this thread on another site but it gives insight as to what is going on poly count wise from consoles and PC games.

nsider2.com/forums/Polygon-Count … 65375.html