Is there a trivial way to convert a highly detailed model to a lowpoly model suitable for games? I'm not a modeler but I'd like to make my own models myself.

I use Blender, and it has an automatic Decimate function that allows you to specify the exact number of triangles you want to reduce to, and will give you a realtime preview of the mesh as you reduce. There are some simple requirements for being able to do this (ie, the mesh can't have any degenerate faces) but otherwise it's pretty simple.

You may want to use something like Decimate to reduce to a certain level, then do manual edge and face removal from there for the best visual results, since not all meshes reduce to perfect low-poly models, and an attentive artist can usually do better. Especially with joints and seams.

Do the professional (and non professional) modelers that create models for games first create a high poly model and then create a low poly from scratch or they use some other techniques?

There is no set way of doing it, at any level.

Some people prefer to start with the high-poly model and build a low-poly "shell" around it.
Some people prefer to start with a low-poly model to get the general form down, then transform it into a high-poly model.

I don't believe, however, that many (if any) professionals would use an automatic function to reduce a high-poly model to the low-poly form. It nearly always results in a model that a)deforms poorly (looks bad when animated), b)has messy UV coordinates (wastes pixel space and makes it hard to texture cleanly), and c)has a poor silhouette (looks jagged).

No way of doing this, eh? And as you lower the number of polys, jagged edges become nearly impossible to get rid of entirely.

How much time does it normally take to create a model for instance for Doom 3 (a pair a low and high poly model) + animation?

How often developers use motion capture to animate 3d models for games?

Quote:Original post by overflowed_
No way of doing this, eh? And as you lower the number of polys, jagged edges become nearly impossible to get rid of entirely.

Do what? Take a simple, overly dense, poorly UV mapped model with hardly any lighting and reduce it to a simple, still wasteful, less accurate, poorly UV mapped model with hardly any lighting and absolutely no chance of being properly animated?

The point of that algorithm in particular is to reduce the density of a mesh when it reaches a point far enough from the viewer that none of the above actually matters, because it can't be seen. That is a very different situation than the making of the original low poly model.

Do you think you could, for example, take an even denser form of that mesh and use an algorithm to come up with the model on the far left, with all of the evenly spaced and lined up quads? If you can't, then that's exactly why professional companies do their low-poly models by hand.

And if you can you should patent it and make a ton of money selling it to every game company around. Dead serious.


Unfortunately Nameless I really can't give a solid answer to those questions. It depends a lot on the studio, and what sort of game they're doing. Motion capture is very widespread, but you can't very well capture a dragon, really. Models can easily take days (more than a week, sometimes), depending on the complexity and the level of quality being aimed for. This is, however, greatly being counteracted by the arrival of better modelling tools such as ZBrush, which can cut down the time it takes to produce a high-poly model by tremendous amounts.

Hi

a few years ago I modelled a few playermodels for a Halflife modification and we had a limit of ~1500 triangles since HL didnt use any advanced rendering functions

The way we modelled was we did a pre calculation on how many triangles to use for which body part * this is something that requires a little bit experience with animations and so on*

I usually imported the Halflife biped to get the exact measurements for a player
and started from the feet upwards
i modelled one leg made it nearly perfect detached it from the biped cloned and mirrored it for the other leg

then the torso

then the arms *same as the legs but the fingers need additional polys which can be a pain to get the proportions right*

thenshoulders *very imported part because of shoulder-arm links for animation since we had no animation blending*

and the head where we spend around 30-40 triangles to make it look detailed


modern games use up to 5000 triangles i have read although i wonder where they waste all these triangles the model details on a model can be achieved with proper shading for 5000 triangles i would really expect a little bit more details