High-Poly to Low-Poly Tools

Neji

I've recently started to try my hand at high poly modelling after learning how to model in 3DS Max. I've watched some of the 'Old Damaged Pillar' tutorial from eat3D and it seems that the instructor uses ProOptimizer (Pilycruncher).

From what I've seen myself it doesn't create the cleanest meshes (unless I'm using it wrong, which is a distinct possibility ) so how do other people go from a model with millions of polys to a low poly version?

Is there a lot of manual work in producing high quality low poly versions of meshes or are there decent tools to do it mostly automatic?

cryrid

It depends on what you're looking for.

If you want to get a 15million polygon model down to a lower number for the sake of importing it into a baking app, that's when things like ProOptimizer and Decimation Master come in handy (clean topology shouldn't be too important, just the surface detail).

If you're looking for a low-poly mesh you can use in a game, you might be better off re-toplogizing it.

Mark Dygert

Yep polycruncher works on triangles so it makes a mess of nicely quaded meshes. Which if the topology doesn't matter then its a good tool to use. I don't think there is a quad based optimizer in any app, anywhere. If there is I would love to.

Export the lowest sub-division and clean it up.

Depending on how it was built Zbrush might be a good option. 3DCoat has some cool retopo tools.

Build a new low poly from scratch using tools like WrapIt or MaxRetopo or Topogun. I personally really really like WrapIt, amazing little script. It has a one button "build quad mesh" feature. The branch painting and extrude edge/snap to surface is great, saves a ton of time. MaxRetopo is free, doesn't have any feature other than conform but works pretty well too.

Neji

I'm basically interested in taking high-poly meshes and making them low-poly for use in game engines, like UDK. So for this, I should either re-create the low-poly version with something like WrapIt or just clean it-up manually?

Thanks for the replies and links, I'll take a look at some of the software suggested.

Mark Dygert

Depends on what its for. If its a deforming mesh like a character then you want to retopo it in some way so you pass on a nicely quad-ed mesh to the next guy. If its a static prop that is semi-organic like a rock or rubble and doesn't deform then its probably ok to crunch the hell out of it and clean it up a little, no one else will have to work on the mesh. If its a hard surface like a sci-fi metal corridor then you should probably make it quads because tris will more than likely give you smoothing errors and wonky shadows.

raul

The way its done on the eat3d tutorial is cool and fast. But is nothing but a foundation to keep building on. If i were you, stick to the basics and build your own low poly by altering the base mesh ( that way you know that it will wrap around your model )

All low poly is, is a low res cage that wraps around the high poly mesh. Learning how to make one properly is hard. But worth the time.

Neji

Depends on what its for.

At the moment, it's will be for environments, so buildings, rocks, trees etc.

All low poly is, is a low res cage that wraps around the high poly mesh. Learning how to make one properly is hard. But worth the time.

Is there any tutorials around that specifically deal with this? I've watched a couple of other tutorials aswell but they didn't really go into depth about creating the 'cage'.

DOG-GY

The cage IS the low poly model. For environments I wouldn't recommend any kind of automatic optimization. You won't get the results you want. You'd want to pull those assets into a program and start making low poly models over them.

CheeseOnToast

Some software is better suited than others for the job, if you go the manual route. Maya and Max can both do it, but it's easier to do in Topogun, 3dCoat or even Silo.

Neji

DOG-GY wrote: »

The cage IS the low poly model.

Yea, I get that. The problem is, I'm not sure about how best to create the low-poly model. Obviously it will differ for each object but how low can you go to keep the details of the high version? That's why some tutorials or something specifically about this would be good.

I guess it can be a case of trial and improvement though?

Mark Dygert

Neji wrote: »

Yea, I get that. The problem is, I'm not sure about how best to create the low-poly model.

This is like 75-95% of the job and up until a few years ago it was the only way to make models, its a skill that is still 100% required.

Neji wrote: »

Obviously it will differ for each object but how low can you go to keep the details of the high version? That's why some tutorials or something specifically about this would be good.

This also depends on a lot of factors.
The type of game.
The hardware it will run on.
What kind of shaders and maps can be used.
How close the player will get to the object.
The angle at which it will be viewed.
The objects importance and scale in the scene.

There aren't set numbers across the board for these things, sometimes ranges for specific scenarios can be given but often even in a production environment its a bit of trail and error and optimizing at a later date.

Another thing to keep in mind is that most engines/hardware can push a lot of triangles but where it bottle necks is in the shaders and textures. If there's any place you need to be wise about its usage its the texture space.

A good rule of thumb is to use as little as you can without needlessly taking up too much time optimizing or making your workflow hell just save a few tris.

Neji wrote: »

I guess it can be a case of trial and improvement though?

Yep. It's all about context, trial and error. When you post an object for critique people will want to know the context otherwise they can't offer any solid advice about the technical side of things. It's probably best to set some limits for yourself before you start. Pick some hardware, find out how many tris it can push, decide how many of those would be reasonable for your object and go from there.

Neji

Thanks. So trial and error it is!

Kewop Decam

Yup, what Vig said.

"the only way to get it right, is to get it wrong!" -Ed, Edd, and Eddy

Yamo

Neji wrote: »

Yea, I get that. The problem is, I'm not sure about how best to create the low-poly model. Obviously it will differ for each object but how low can you go to keep the details of the high version? That's why some tutorials or something specifically about this would be good.

I guess it can be a case of trial and improvement though?

In Zbrush I like to take my hi-poly mesh and actually paint my low-poly mesh onto it like a texture. Then, starting with a zsphere, make a new mesh based on the grid I just laid out onto the hi-poly

spetch

if you're running low on time expecially in a production atmosphere I suggest mesh lab's Quadratic Edge Collapse Detection. It does a pretty good job of averging the tris. But it is still tris...and it can sometimes turn out messy where you have to manually tweak the geo afterwards.

Here's a good run down off the splash damage forums-
http://www.splashdamage.com/forums/showthread.php?p=213472

Agree with Cheese on toast- Topogun and 3dcoat work wonders for retopology. I dont use max graphite modeling tools anymore.