This is the (mostly) completed mesh of a model based on one of my dogs. This is the first model I have built on my own initiative – i.e. not following a tutorial.
I tried to upload to sketchfab but I don’t really know what I’m doing with exporting and things didn’t turn out right, so hopefully these screen grabs give a good idea. There are some details I want to add around the eyes and the eyes themselves, but otherwise I think there is enough detail to suffice for a reasonably realistic looking model once I get some fur and textures put onto it, which is my next step. Right now it is about 35k polys. I am not sure how much the fur will add – I expect a lot. I probably could decimate this since much of it will be hidden anyway, although it is a short fur dog.
I don’t know much about shading at the moment, so I have just slapped on some basic stuff so that it’s not matte gray. My next step is to learn how to make fur and then do that. With that done, I’ll move on to rigging and animation. If this turns out satisfactory, that should cover all the skills I will need to start developing assets for my own project (which will include different kinds of critters, but not dogs necessarily)
Being my first model ever, this took me upwards of twenty hours. Most of my time was spent fixing up bad geometry, experimenting, and playing around with different edge flows to try and see what would allow the most movement without pinching since I want to animate this later.
I made this in Maya 2017.
Please share any thoughts, criticisms, suggestions, etc. I really appreciate them!
I think this is a really good start for a first model. I would suggest to give the area around the eyes another go, that’s one of the most important and still looks off to me. And your polygon density really is excessively high for that kind of asset. You could and should make a version with much less polies and bake the details of your highpoly version into a tangentspace normal map *). I suggest to try Substance Painter 2 for your texturing needs. It’s reasonably easy to learn and can do almost all the baking for you. I would suggest to keep rigging and animating as the very last steps because you’ll likely modify the mesh many more times during creating the lowpoly version, unwrapping and baking. Fur is tricky, but I have no advice for you, since I never had to tackle that myself.
*) it’s not so much that it makes a big difference for the GPU whether you have that one model at 5k or 50k triangles, but if you have 10 of them on screen and it’s 50k vs 500k, it can start to add up. You just don’t want to be wasteful with your resources in general. I don’t have hands-on experience with skinned mesh-renderers, but the impact of having to calculate skeletal deformations during animations, with many vertices being influenced by more than one bone at the same time, could have a bigger effect on CPU time, than just rendering that amount of polygons on the GPU.
Whether your terrain mesh has 500k triangles or 2 millions, I wouldn’t care much about as long as your framerate is what you want it to be, because you likely only ever will have one terrain and won’t animate it. That’s a known quantity that won’t change. But down the line you might change your mind on how many animals you want to be able to have on screen at the same time etc… Better to stay within reasonable limits from the start.
When building the lowpoly focus on keeping the curves in the silhouttes of your model smooth enough, everything that you never see as a silhoutte and doesn’t get deformed heavily during animation can be pretty low poly because the normal map will preserve all the nuanced detail. E.g. the sides of the torso can be pretty low res, the areas around bendable joints should have higher resolution in comparison.
One thing to consider, and then see if you still stick with that first sentence – my project is a hunting game, which would typically only have two animals one screen at once. Yourself and another critter. Sometimes there might be more, but that would likely be from a far distance and thus low poly versions could be used. In that case, don’t you think it is reasonable to have some relatively high poly models?
Thanks a lot for the good advice. I appreciate the mentioning of terrain – that is where I want to put the most resolution as besides the sparse critters the terrain will pretty much be the only other thing in the game (not counting running scripts of course).
To be honest I don’t know what baking the details exactly entails. It is one of those things I am going to need to learn along the way. I think I need to complete some basic texturing primers first though, as that is prerequisite to that sort of thing?
And, one question for you or anybody who knows a thing about modeling:
If I want to reduce the amount of resolution in places like the sides, but keep a decent amount around the nose, eyes, etc., how do you suggest rerouting those edges? I couldn’t seem to find many inconspicuous places to merge a bunch of vertices – I mean, I didn’t want a phalanx of kites right along the shoulders. Perhaps I could route those edges down towards the belly and maybe loop them? I guess I ought to just try that.
But maybe I don’t even need so many around the nose.
Even if you’ll ever only have 2 animals on screen at a time, you’re still throwing polygons away in a way that doesn’t get you any tangible benefit. Once you’ve familiarized yourself with modern texturing workflows you’ll easily see how there are areas on your mesh where the current amount of polies is real overkill and others where you could add more to actually get a visual improvement.
I’m guessing you’re using subdivision surface modeling? With that technique edgeflow/topology matters a lot. For a low-poly model intended for games and the usual highpoly to lowpoly baking workflow, topology doesn’t matter nearly as much. You’ll still want to avoid a few pitfalls related to normal map baking and deformation during animations, but you don’t need to think about the same things you need to think about right now with an SDS workflow. Typically you’d leave your model right there as it is topology-wise, and build a new lowpoly model to actually use in the game. The current would then be your highpoly model to bake textures from.
Search term that will help you: “retopo” or “retopologize”. I’m sure Maya has some tools for that.
[EDIT:]
You definitely don’t for the lowpoly ingame model. For the highpoly if you were to start sculpting you might want to add a couple hundred thousands right there to model the tiny bumps on the nose. You could also use a pbr material in Substance Painter that already has a suitable texture. There’s usually more than one way to achieve things. If you use the workflow of baking highpoly to lowpoly, then your highpoly can have as many millions of polygons as you want, it won’t affect the rendering performance in the game once it’s turned into a normalmap.
If you want to get good at this, you should open an account there and seek out feedback there. You’ll probably receive much better answers than you could here.
Thanks so much. I think the most powerful kind of advice a noob can get is a good resource. I’d browsed Polycounts art showcases before, and I was left intimidated and feeling beaten. Haha. But if you can’t beat them, join them, right?
I think before I ask for any more time from kind strangers on the internet, I need to cover all the basics with some more tutorials. I guess I just wanted to make sure the way I was working was total crap, lol.
The main thing to keep in mind, is you want to keep quads as much as possible. A couple triangles are alright in non deformable areas, but usually you’ll want to re-route a row of edges/polys and keep quads rather than collapsing down to a fan/pole. And stay away from polys with more than 4 edges = ngons.
Pinterest has some good modeling tips/techniques to keep models with most (95%) quads. Look up poly flow or quad modeling, that should pull up some good reference for you.
Also another important point to always keep in mind while modeling - follow the muscle flow as much as possible for organic models. This will help immensely when you finally get the rigging and animating. Several references available for modeling for best animation results.
Last suggestion - Keep modeling new models! Consider putting this model aside and creating another completely different model totally from scratch. This will help you get better faster and then, go further with the second model. Finalize the modeling phase and ready it for UVs and unwrap it. After you’ve created 3-5 additional models you will look at this dog model and (even though pretty good for first time model) you will look at it and either not like it at all or want to start over from scratch because you have learned so much creating the last 3-5 models.
Thanks a lot. Reminds me some advice I read somewhere – maybe it was here on this forum but I really don’t remember – but the idea is that in the beginning going for quantity over quality is the fastest way to learn.
But yeah, I totally agree. I started poking at that model this morning trying to get things perfect when I realized, hey, I’m not learning anything new doing this. I wanted to learn to do fur next, but I realized pretty quickly that was a little too advanced for where I am right now, so I am going to how to do UV layouts in Maya right now. Perhaps after this tutorial I will have another go around with a different model like you suggest – I think a human with a detailed face.
Although I am having a great time and learning a lot recently, sometimes I wonder if I am on the most efficient path. My end goal is to be able to make assets for a game entirely myself and from scratch, but the problem is that there are so many tools its kind of hard to know where to go.
Maya seems like it will be my all-in-one powerhouse, but I know that I will need something like Photoshop as well for creating textures. As I have already dropped money on a new computer, lessons from Pluralsight, a Maya subscription, its not likely I’ll convince my wife at this point to allow me to purchase any other expensive programs, so I kind of need to just learn everything I can do with just Maya and Photoshop (which I don’t have yet, but have convinced the wife to let me get later).
TLDR – A bunch of random thoughts with no coherent point. Disregard.
Check out a substance painter trial version and then maybe put the steam version on your wishlist to get notified during 50% off sales. If you can get to grips with its texturing workflow I see little reason to even need Photoshop for making realistic PBR textures. You still will want and need a tool like Photoshop for some things like UI art assets or tweaking photos to use as masks or textures in substance painter, but in terms of what gives you a fully textured and usable PBR asset the fastest, you’ll get orders of magnitude more utility out of substance painter imho. If you aren’t used to Photoshop yet, you might as well give Affinity Photo a try first if they have a trial version. It’s much cheaper I believe.
I think I’ll do that about substance painter. It seems like learning the most bang for your buck programs first will be my best option. I take it Substance painter is just a one time purchase deal then.
So I just watched a basic UV mapping tutorial and a retopology in maya tutorial back to back at 2.0x speed. Todays my day off so I have to make the most of it. I didn’t follow along like I usually would, I just wanted to get the basic ideas. I think I understand the process now, and it doesn’t seem terribly daunting either.
In a nutshell, I can make a very nice looking mesh, edge flow doesn’t need to be perfect, with all the detail I can muster. From that masterpiece, I will use the retopo techniques I learned to make a more animation and GPU friendly model, and the smaller details I can bake from the masterpiece for a diffuse and normal map (and the other maps too, I suppose, though I don’t know yet what they all do.)
That all seems pretty doable, though I will need some practice getting used to the retopo tools in Maya.
After that, I’ll move over to Substance Painter and start learning how to do that. Since I don’t want to buy that right away, I will just do the baking in Maya. Perhaps Substance Painter might do that better, I don’t know, but I’ll just stick with what I have for now.
I think, like suggested, I’ll save rigging and animation for the last thing I learn how to do. Perhaps hair and fur as well – though it might be worth learning that early on since I’m making a game about animals.
Although learning the built in software hair/fur/cloth/particles is really nice to know for future use in 3D and cinematics and other stuff, the hair/fur/cloth/particles in each software (maya, 3D Max, others) do not transfer into Unity.
Also you will find hair/fur/cloth/particles in 3D are much more robust in the 3D packages than they are in game engine, though Unity particles have made leaps and bounds improvements since legacy particles, and even 1st gen Shuriken particles.
This does not include the low poly hair techniques commonly used in Final Fantasy games, Shadow of Colossus and other games. That technique is a union of 3D software, advanced rigging, and programming/physics system set up in engine/Unity.
Wow. That’s good to know. I won’t prioritize learnings Maya’s nParticle systems then. I wonder how best to go about simulating animal fur/feathers?
I suppose really nice textures and a lot of time spent making small tweaks in animation should suffice. For instance, an idle animation for a bird soaring in flight could involve several similar iterations in which lots of different individual feathers are tweaked to look like they are being affected by wind turbulence. This seems like it would be tedious, and perhaps solved by some advanced programming algorithms, but certainly doable.
I take it the programming/physics involved means applying custom solutions for the hair to be animated? Something that is likely way beyond my scope, but interesting nonetheless.
The modelling is rough but the reason why you’ll go much further than 90% of people is because you understand what you are looking at. You observe the smaller details and then replicate them.
Most wannabe artists don’t. They just get what they think they’re seeing and make some sort of crappy lump.
In your case you have the forms but not the experience, so the only thing you actually need to do is just do a lot more. There isn’t much we can teach you in this most important stage. Good news is, you’re already cracking the most important stage: observation.
The rendering and the FX, none of that is important without the underlying model being good, and you’re good at that.
About thinking more along the lines of cinema versus games, that’s probably because the tutorials I am following are not geared towards gaming in general, just learning how to use Maya. So sometimes I don’t know if what I am learning will be specific to my task or not, but slowly with the help of this forum and other resources I am. I would love to find some tutorials that just cover specifically what I am interested in and nothing else, but right now Pluralsight has the most and highest quality tutorials I’ve found.
In a couple months or however long come back here and post and I can give you some helpful input on rigging feathers, wings and such for games and with realism in mind. Not a beginner topic - and hippo said it - keep at it!
You guys are awesome. Thanks for giving so much to the community.
Here’s a snippet of what I’ve been working on. I may finish the high res mesh tomorrow. It’s easy to get carried away, so I’ll be mindful not to strive for perfection. The more important things to learn at this point are retopoing the mesh and doing so in a good way to be ready for texturing and animation.
Anyway, here’s a pic. It’s a fantasy character from my imagination. Sort of a cross between Dracula, a werewolf, and he’ll have big wings so I can work on feathering. Obviously it’s a work in progress, but after lots of experimentation I finally got the general look I was going for.
My method so far has been this: Follow along a tutorial for creating a general human head. I made the body myself but used a human proportions diagram to build a scale template so it would be as perfect as possible. I will make the legs resemble a wolfs and probably elongate the hands a bit though, but I wanted to start from correct proportions so it would look believable.
Even though the lines don’t matter too much, I took a lot of time to get all my edges running as efficiently as possible just for practice. Then I subdivided to about 500k to get the general shapes I wanted, then subdivided again to 2 mil to get little fine cuts to look like hair. I don’t think I’ll need to go any higher than this.
