Decimation Master an acceptable shortcut?
polycounter lvl 8
Hi, all. I was wondering what the general consensus is - if there is one - on using shortcuts like decimation master for finished pieces.
Specifically, I'm considering simply using zbrush's decimation master plugin to get the low poly teeth geometry for my character. Decimation master is likely to give me a better optimized mesh than if I retopologized by hand. The only downside I can see is that the topology will look ugly.
Is this considered unprofessional?
Specifically, I'm considering simply using zbrush's decimation master plugin to get the low poly teeth geometry for my character. Decimation master is likely to give me a better optimized mesh than if I retopologized by hand. The only downside I can see is that the topology will look ugly.
Is this considered unprofessional?
Replies
Regardless, I find that decimation master is great at
1 - achieving a specific poly count
2 - making the most of those polys to preserve all the details.
I agree. Although I've just started getting into sculpting and baking and such, I find it's much easier to use something like Decimation Master to get a quick low poly and depending on how messy it is just use it or clean it up a bit.
The real question is how do they look with a normal map applied in the game engine, and doing the uv map/textures/bakes.
As for your example, I'd go for an hybrid solution (if you REALLY want to use decimation master to "speed up" the process) and do some geometry tweaks to avoid stretched/useless triangles.
Whatever works !
That's how I normally approach things like this.
I'm going to stick with hand-retopo for this particular project, but I'm gonna stay open to decimation master in the future. The trick is finding that threshhold where it's better to retopo than to decimate.
Listen, I understand we're all trying to get the 'quickest' things to get with it done, but tell me this, how are you going to unwrap a decimated model? Or have it bake correctly? Or break is apart if the scene requires a specific breaking of a model? There is a reason topology is important, even on low-poly model, extruding, unwrapping, extracting, bending, animation, all this, even on static assets can be strong points if you have the right topology.
The only time I would say such quick and easy methods are acceptable would be if I could personally see in realtime the decimation of said model, and one that uses a normal based decimation instead of what ProOptimizer and Decimation Master do, which is decimate by vertice threshold.
There is a software, I don't remember the name (Adango or Adagio) which offered a decimation solution, I think it was realtime, and didn't cause any nasty normal smoothing issues, and kept the decimation much cleaner then ZB.
In your picture, I honestly cannot see how that low-poly retopo model is 1.2K vs decimated master which nets in 1.1k. Those teeth shouldn't pass the 500 tri's in my opinion, unless you plan on having Kratos jump and mess around in them, in which case, they need to be higher.
I'd actually like your input on how well it turned out. What would you guys have done in my place?
Now correct me if I'm wrong, but if something needs to shatter in-game, then that would likely be planned in advance, yes? Same for animation, etc. I obviously would never decimate something that needs to deform. I guess what I'm saying is - should we really build everything in such a way that it could deform/shatter if it needed to, even if we don't think it will? That seems overly cautious.
You're right about the teeth. I'm within poly budget for this character and I hate to strip out detail, but I'll likely go back and reduce the poly-count on his teeth.
I think that's a perfectly acceptable use of Decimation Master - however, you could still have gone in and cleaned up further. But it's certainly a good use of a tool.
It really is not a BAD approach to working. Take for example POV arms (first person arms) for a shooter. You could definitely have a fully uniformed/clothed arm and have it decimated for sure. Though you might wonder where do we draw the line for proper topology?
Simply all hands, feet, faces, shoulders must carry proper edge loops. Though if the arms between elbows, wrists, knees carry only clothed parts. The looping is not necessary when flowing from a region to the next. Though the loop must exist in the the deformed joint>to>joint portion. Or in other ways, you can get away with topology murder by having enough points in a tight bending region (elbow for example.) and still have proper deformation and no one can notice.
- My example of this
>
It can work in a pinch if you need a dirty test bake going but we retopo all our assets properly eventually.
Decimation master is great for saving memory on bakes but I am not sure it's a good production shortcut for game meshes.
Like my big hairy balls told me, I can literally see a few big quads (Krogans) and several small triangles (Asari) on that model that could cut down the polycount (Polycount...erm) in that model!
Listen, I'll be frank, while that is good, ALL that work could be done by your normal map. If you wanted a better 'deformation' on the mesh, use a higher polycount, don't make it half-way! That just looks tacky and horrible. YES IT WILL WORK! I don't doubt that, but it's neither low-poly, nor medium poly, it's kinda in the twilight part of being stuck in a quantum of state of Low-Poly and Medium-Poly, which can be kinda confuzzling. It doesn't save resources as best as it could, nor does it bring out the 'pop' you're hoping if you apply a normal map to it.
If you only used a Diffuse map, then yes, something like that would be great...
Also, about the part of Animation Loops, don't cry victory just yet. A single Turbosmooth in Max will send your model in a nose dive for performance like Al-Queda Airlines, most game engines either use Morphs or Cache the Vertices from the bone animation. Remember, bones are moving vertices and each point adds to the stress.
CrazyMatt:
Nice approach mate, but my cynicism alarm went off like a flaming hernia in mid-July, while a firework was stuck up my arse on heatwave.
That model will be, for a lack of better terms, a bitch to skin. I personally wouldn't be able to see any quality work being done on skinning that...thing. It's good looking and will work, I know it will work, but I cannot see any quality work being done it since it's too erratic. I also have a hard time imagining how you will enable modders to texture it.
Also, as I said previously, it's in that mushy middle, my biggest problem is that someone can literally retopo that part, clean and with the deformation in under 10 minutes, UV'ed and all as well skinned, and save up a bunch of polies as if it was nothing.
I'm not talking about optimization, but something on which as an asset, I can reuse later by simply pulling and relaxing for another pair of arms because it's clean.
Again, it look good and from what I see it will work, but it's just so unnecessarily cluttered that I'm not sure if you saved any time with the extra work that needs to be done for Skinning or Texturing. If could explain those two parts, it would be great.
I'm not sure I see what the problem is with skinning or texturing CrazyMatt's arms. Skinning is simply a matter of painting weights on vertices. If you have enough vertices, topology doesn't even affect it. You can see that he properly retopo'd the wrist and fingers, since they would need more precise deformation. Uving can be done prior to decimation. As you can see, I had no trouble uving - or texturing - my character's blade.
The problem area on my model was not the surface of the blade, but the edge. The edge is so crooked and misshapen that I wasn't entirely sure how to approach retopologizing it while preserving the silhouette. I COULD have just powered through it, but I found it easier in this case to simply decimate. I went in afterwards and cleaned up some excess geometry. While I could try to optimize it fully, there isn't much point. I would only get rid of ~100 more tris, which is entirely negligible.
I'm not sure what you're getting at with turbosmoothing or bones. Decimation occurs after sculpting, so there's no reason to turbosmooth. As for more verts-per-bone requiring more memory...yes. Higher poly models will require more memory. Decimation isn't about just throwing excess polys into a model. It's about reducing the polycount to a desired number while retaining form, without time-consuming re topology.
Those first person hands are way overbudget for the amount of detail (or lack there of)...
Perhaps you guys have a slightly misguided idea of what decimation master should and shouldn't be used for.
100 tris are actually a lot. I mean what if you would do the same thing for several different parts?
Also I think ace-angel is talking about skinning in 3ds max. Selecting each of those vertices and setting the weight can get pretty tiresome(as opposed to painting the weights in maya).
When I read something like this, I ask myself... "Why for this political corrective attitude?"
There is nothing wrong with experimentation, and if it proves successful for the purposes it was intended for. Then all more gravy to the mash potatoes.:)
I am not sure what you're trying to say here...
Anyway I just wanted to point out that this application of decimation master will not work for skinned characters in a production environment. It's a handy tool for managing high resolution meshes it's not really meant to be an alternative to making proper real-time meshes. It may work in a pinch for static objects but even then, only in a pinch.
For the record, I have worked in a production environment. I recently handed this model off to my friend who'll rig it while I texture it. I haven't heard any complaints from him (besides the fact that it's in an A-pose).
@Goraaz Yes 100 tris would be a lot for something like a Dead Rising zombie, where there'll be hundreds on-screen at once. This isn't that case, however. My character is meant to be comparable to Pyramid Head or the resident evil Nemesis.
I'm pretty sure you don't have to skin in max by selecting individual vertices and tweaking values. I've never had to skin in Max, but that sounds so bass-ackwards that there's GOT to be an easier way.