Cutting Down Polygon Count On Characters

[Domarius]

3DS Max has an MD5 importer. I downloaded it already. Will try it out in a sec.

[SneaksieDave]

From the other thread: I looked briefly at the md5 importer for Blender, and don't have much hope for it. Even if I did get it to work, I'd be willing to bet it doesn't preserve UV maps. So, to any modelers who might know: is that really so tragic as it sounds? Can the old UV map just be "reapplied" to a modified and exported model?

[Domarius]

We have the original MBs for all of the models though, so we don't need any MD5mesh importers *hits head*

 

It's not impossible, just a lot of work. Everything will have to be re-rigged and re-UV-mapped.

 

And at the same time, if we want to make decent thief levels, something has to be done about it.

 

BTW, the 3DS Max MD5 importer works REALLY well, bones and rigging intact, so, if we ever need it... there it is.

 

Well, we'll get squawk's template model pack on the weekend, and then we can all take a breath and think about where we want to go with the poly count of everything we have.

 

I suppose if squawk's templates are made lower poly, now is the time to do it, before anything is made from them. Maybe the switch to mitten hands can be a part of that.

[New Horizon]

You guys should confer with ascottk about this stuff, he's the guy working through all the setup glitches...so it would be good to run this stuff by him...as it's likely going to cause him more work.

[OrbWeaver]

From the other thread: I looked briefly at the md5 importer for Blender, and don't have much hope for it. Even if I did get it to work, I'd be willing to bet it doesn't preserve UV maps. So, to any modelers who might know: is that really so tragic as it sounds? Can the old UV map just be "reapplied" to a modified and exported model?

 

I am fairly sure it does import UV maps actually. I did use it once, but I haven't done any modelling for a while so I can't remember how easy it was to get working.

[Domarius]

You guys should confer with ascottk about this stuff, he's the guy working through all the setup glitches...so it would be good to run this stuff by him...as it's likely going to cause him more work.

Yeah that goes without saying. He's the only one really technically capable of these sorts of things with any decent speed. We're just talking about it is all. He's not here right now.

[Springheel]

And at the same time, if we want to make decent thief levels, something has to be done about it.

 

Something has to be done about *what* exactly? Where did all this panic suddenly come from? We've been working with the AI for ages now, and--while they may not be perfectly optimized--they're a reasonable polycount and they're not causing anyone's system to start crawling (outside of processing stimuli, that is).

 

We've already got plans to create simplified shadowmeshes for them, which is really the biggest performance hit, so unless there's some new development I'm not aware of, people need to chill out.

[Domarius]

Heh... yeah.

 

Spar will have something to say about this.

 

There are ways to measure what is the most processor intensive part of the program, and then once you find out those things you focus on optimising that, and nothing else. Very old rule of optimisation. How could I forget old Michael Abrash.

 

For the record, I wasn't panicing, just going with the flow. 3am gives me a very short term memory too. You can see there I was talking about md5s, when we already have MBs. I also forgot about the shadow map thing.

[oDDity]

IS this really worth the extra work, I mean, even if you cut them all down by a third (which you can't) all it means is that you can have 4 characters on screen for the same polygon price as three of the old ones.

Also, by the time this is released, which will be 2008, you'd imagine any gamer will have a computer far exceeding the standards of dooms original recommended system specs.

[Domarius]

That's another way to look at it. Also along the lines of Michael's optimisation philosophy.

[PinkDot]

We've already got plans to create simplified shadowmeshes for them

Does it mean rerigging models and reexporting animations? Isn't it better to make all new characters with shadow meshes already in?

 

Also, by the time this is released, which will be 2008, you'd imagine any gamer will have a computer far exceeding the standards of dooms original recommended system specs.

But Doom recomended system specs are obsolete in relation to our mod... Doom3 works smoothly for me - our unfinshed and small maps not... (or nearly not)

[oDDity]

Thinking about it, the ratio isn't even that good, because it's not just about pure poly count, its about materials and animations too, so it won't follow that the extra character made from the polys cut off the three old ones will mean that you end up with the same net performance.

[ascottk]

We have the original MBs for all of the models though, so we don't need any MD5mesh importers *hits head*

 

It's not impossible, just a lot of work. Everything will have to be re-rigged and re-UV-mapped.

The uv maps are there with the max importer but the meshes are combined. It's easy to separate the meshes because the elements are there. You only have to select which elements you want to separate.

 

BTW, the 3DS Max MD5 importer works REALLY well, bones and rigging intact, so, if we ever need it... there it is.

I use it all the time & I'm surprised it works this well. I imported one of the zombies for Solis to animate. It's the md5 exporters I had issues with.

 

We've already got plans to create simplified shadowmeshes for them, which is really the biggest performance hit, so unless there's some new development I'm not aware of, people need to chill out.

Shadowmeshes should be pretty easy. Make a copy of the meshes in maya, combine them, lower the poly count (I tried this & maya seems to do this really well), reassign them with a shadow material, rig them, import to d3, then add no shadows to the chartacters' materials.

 

Actually it'd be time consuming

[OrbWeaver]

You wouldn't necessarily need to create a brand new shadowmesh for each AI, providing they are fairly similar in size and shape you could use one generic shadowmesh for quite a number of AI.

[SneaksieDave]

Hmm. Mixed feelings about this now. Some good points raised here, both sides of it.

[Domarius]

You wouldn't necessarily need to create a brand new shadowmesh for each AI, providing they are fairly similar in size and shape you could use one generic shadowmesh for quite a number of AI.

So true - nobody's going to look for belt buckles in the shadow.

 

You'd make a basic humanoid shape, maybe seperate ones for male and female, and then make small changes like adding low poly pauldrons. And the same shadowmesh helmet, would probably work for every helmet in the game, same for female dresses.