Low- & High-poly modeling?

[heavyness]

i'll model something high, but keep in mind i'm trying to use the least amount of polys as possible.

i would rather go back and rebuild a character's shoulder area then waste time when i'm first trying to flush out the character.

but like others said, it's your own decision and workflow.

[RALawrence]

My process varies depending on the type of object. Lately I've been working on sharp rocks and cliffs, and I've found that I get the best results by quickly modeling a lot of low-poly shapes, throwing them together in a visually appealing clump, then taking that into Zbrush for sculpting. Once finished in Zbrush, I export the highest subdivision AND the second or third level of subdivision back into Max, where I use Polyboost to model my game mesh on top of the lower subdivision exported mesh. Then I lay out my UVs and bake from the high subdivision exported mesh.

However, there are times when I have to create a piece of architecture the fit into a set that other people are working on. In a case like that, it's important to model the game res mesh first, fit it into place, make sure it works well with the rest of the set and has workable collision. Then the high poly

Oftentimes on hard artificial objects I will make a medium-poly model from my low-poly after I've laid out UV's. I bake my normals and AO from the medium instead of the low-poly model. The extra loops that I put in at edges and other problem areas help me to eliminate unsightly errors in my baked maps (i.e., "the waves").

[Talon]

With your hard surface process, surely the mid-poly mesh ends up gathering incorrect normal data as the normals of it and the low poly will be so different?

[RALawrence]

To clarify, I don't change the shape of the model at all when I derive a medium-poly version. I simply add edges at key spots.

Check out the attached pic. The medium poly in this example had only two extra slices, but results in a cleaner map.

[Talon]

Ah, gotcha

[Flinet]

Nice! So if i'm getting this right;
You made a medium poly mesh and took the normalmap and AO from the highpoly mesh.
After that you took the normalmap and AO and put it on the lowpoly mesh?

If im right I have to ask, how did you manage to have the same UV's on the medium- and lowpoly mesh? Did you do the mediumpoly mesh first and then just deleted every second line to keep the UV?

[retardedmonkey]

Originally Posted by RALawrence

To clarify, I don't change the shape of the model at all when I derive a medium-poly version. I simply add edges at key spots.

Check out the attached pic. The medium poly in this example had only two extra slices, but results in a cleaner map.

Those "waves" you get from the HP should be because of the cage in the projection options. If you pull the top and bottoms tighter to the faces of the LP you will see that it doesn't give those wavey edges.

Looks like you found a way around it, but figured I'd let you know in case you didn't already know :P

[RALawrence]

Filnet, I UV map the low poly first, then add edges where needed to form the medium poly.

Retardedmonkey, for an object as simple as this I think you are right about the cage. But how do you handle more complex objects? For example, say I had something roughly cylindrical but the top has a lot of detail that I need to capture. If I bring down the top of the cage, the rays end up intersecting the cap portion of the high poly at very steep angles, resulting in a distorted normal. Do you prefer to detach the top faces and bake those separately with their own cage?

[retardedmonkey]

Yeah how I'd do that is I'd either bring it down as close as possible to the geo on top/bottom of the HP. Then I'd just edit the little waves out in photoshop...since usually its just a matter of sampling a color and painting a straight line.

However if I was lazy or couldn't go that route I would do 2 versions of the normals. I would render it out so that all the detail is capture then pull the cage in nice and tight regardless of what it will do to the tops details. Then I just put them together in photoshop. Because the waves are only on the sides of the cylinder so the top won't change when the cage is tighter.

If that doesn't make sense I can make a nice little picture explaining it :P

[Flinet]

Hi guys!

I have modeled this memorystick, i tryed to model the highpoly first and now i have uvmapped the lowpoly. Works fine until i have to make the normalmaps.
Look at the picture:


I use Xsi so ive been using the ultimapper. Ive been computing the "Distance to surface", adjusting tangents, Geometry Approximation and re-uvmapping it. But there will always be seams in the normalmap where the uv-map break from each other.
Anyone that knows how to fix this?

This is generel my biggest issue right now, to make the normals look good where the seams are.

Thanks!

//Jimi