[SOLVED]3DS Max 2011 textures & poly count

[Wyldhunt]

I know very little about modeling in general, so expect me to say things that make no sense. :P

Here's our problem:

We recently updated to the newest version of Max and Poser. Poser has a plug-in that allows us to use our Poser models in Max, which is cool.
We have two problems facing us that we need to figure out before getting our poser models into UDK:

1: Poser uses a very complex material system. The end result is that the 3DS Max model has a lot (50ish) of textures. If I'm understanding the new Max material system, it is all in one channel. There's a huge pile of nodes (Similar to the Material editor in UDK) that controls how the textures get layered and set up so that they show up correctly on the model. We can't have a material with 50 texture calls in UDK.
Q: Is there a way to skin the final texture off of the model as it appears on the model so that we have a single texture file? We need to compress the big texture tree into a single final texture that is based off of how the character actually appears.
I'm hoping that this is easier than it seems to be and someone'll reply with something like "When you export with ActorX, it'll export a UDK friendly texture for you."

2: The models are high poly, which is good for the high poly model side of things. But, it means that we have to go from high poly to low poly. We usually go from low poly to high poly to avoid all of the work of reducing a model.
Q: What are people’s suggestions on the best way to reduce the poly count of a model with the most accuracy? I know I've heard of various software that claims to do this, but I don't know what works best.

[obihb]

1. You can do this by using Render To Texture. The RTT tool can both render directly from an existing model into any UV channel or project from one model to another. I'm not gonna explain the process in detail here but look into that. Max will have some help on there but I'm not sure how detailed it is.

The basic idea is this. If you have a model with 50 UV channels and textures and you want to render that directly into a single texture then apply one more UV channel, 51 I guess. Do a proper single UV layout in that channel. Make that the target channel for RTT tool. It will bake out a map using that channel from all the others. Then you can copy channel 51 to channel 1 and kill the rest, to break it down to a single final channel.

The idea behind projecting is exactly as rendering normal maps accept you render out a diffuse map in stead. You just need 2 models. One with the 50 textures on it and one new one with it's own proper UV layout. Project the 50 channel model to the 1 channel model using the same RTT tool.

2. If you have to do a lot of reduction then it's better to simply build a new low poly model. Max has some awesome tools for this now which is basically derived from the old Poly Boost plugin. The main part is that you can use your high poly model as a base to draw directly on it new polygons. The poly tools allows very fast creating new polygons in this manner.

Your results are much much cleaner and more game friendly than any automatic poly reducer can give you.

If your reduction is not that much then you can simply go in and reduce polygons manually. But this is only really useful if reduction is required at a minimum level. Doing this can be very time consuming on high reduction.

With that said, Max does include a fairly good poly reducing tool but I don't know how well it works on characters. For some stuff it's really good but I can think that for characters it won't be since they are much more precisely modelled with regards to keeping details in certain places.. etc.

It's the Pro Optimizer tool and has various options for various results. Though you can set it up to keep UV seams, for best reductions you're better off loosing them and making new UV's afterwards. But again, for characters it might not be a great idea. You may lose things you don't want to lose and the fact is that auto optimizers are never 100% exactly what you want. Thy're possibly good at creating some LOD's or stuff like that since the model is then far away and can't spot problems on them.

But, again, I would suggest to build a new low poly model rather than reducing a high poly through automatic means. You said you prefer to build the low poly first then the high poly. The fact is the work is the same but by already having the high poly model, making the low poly is much much faster, using poly boost tools of course.

[Wyldhunt]

That sounds like exactly what I was looking for.
Thank you.
And, yeah... Its poly count is about a million right now, so I vote on just drawing the new model over the current one.

Thanks again for the help.

[TMichael]

2: The models are high poly, which is good for the high poly model side of things. But, it means that we have to go from high poly to low poly. We usually go from low poly to high poly to avoid all of the work of reducing a model.
Q: What are people’s suggestions on the best way to reduce the poly count of a model with the most accuracy? I know I've heard of various software that claims to do this, but I don't know what works best.

Hey Wyld,

Even the best poly reduction algorithms are notoriously bad when it comes to the human form. If you have to go from high res to low res, I think the best option would be retopology.

This essentially means building the new topology right over your high res mesh so that you have a near perfect fit, and topology that flows and deforms as you need it to. As Obihb stated above, Max can do this, but there are some excellent tools available which can really help to speed things up. Have a look at 3D Coat's retopology page for some workflow examples (the stroke tool, shown as connecting slices on that page, just blew me away).

I just started working on this guy as a test subject. I call him Mr. Smooth:

[SHOT]http://img254.imageshack.us/img254/3757/mantest.png[/SHOT]

Here I tested the auto-retopology function to find out if it was a feasible option for game characters. I plugged in a poly count of 5000, turned on X symmetry, and did no work on the model after it was generated:

[SHOT]http://img696.imageshack.us/img696/5806/mantest2.png[/SHOT]

Doing the retopology manually would yield the best results. However, after reworking the face and fingers, and maybe adding some extra lines at the joints, this lower res model could probably be mapped, skinned, animated, and ready to rock. Not bad for a quick workflow. Of course, without clothing, hair, and proper textures it would make a poor game model.

[darthviper107]

You don't need Projection to use the Render-to-Texture tool to do this

Lets say you've got one object with lots of different parts that each have a different material applied, so one object is using more than one material

You have to apply a new UV channel and create a new UV map where everything for that object is flattened out.

When you use Render-to-Texture, make sure you set it to use that UV channel, and then just set it to render a new diffuse map. You don't need to project it onto another copy. You only need to use projection if you're trying to render something from one model with different topology to another (like to create a normal map or create textures from a high-poly textured model to a low-poly version).

As for reducing the poly count---do it manually, you'll get the best results, and make sure you think about how it will deform when you animate, really around the joints you don't want it to bend weird.

[Wyldhunt]

Thanks everyone. Lots of help.

[punchface]

Moving from High to Low...

Topogun.

Typically, people in the industry are using this to get low poly models out of work started in Zbrush...

...but I don't think that this HAS to be the case.

punchface