Dand's scrapbook
polycounter lvl 7
Hi there, I've been lurking on polycount for a while and I thought that it would be great if I made a thread/space for myself for all my projects(past/future) so I can keep myself motivated and maybe receive some critiques.
Staring with an older project that I've been neglecting.
I've been working on this guy on and off, for the last couple of months, making little to no progress.
Feels like I always fall off the bandwagon with this one...
Staring with an older project that I've been neglecting.
I've been working on this guy on and off, for the last couple of months, making little to no progress.
Feels like I always fall off the bandwagon with this one...
Replies
A couple of days ago it looked like this
Attempting to blend in a skin shader, hopefully it'll get somewhere better with it.
Oh and uh, critiques please, ty.
I seem to have stumbled upon an issue with his mouth where if I need to add more geometry to it, i might need to go back and remake the whole lowpoly version since it's just a simple nonsplit dynameshed successor.
Front
https://i.imgur.com/A7vC3yn.jpg
3/4
http://puu.sh/fQcfZ/b2fd68c1d1.jpg
Side
http://puu.sh/fQcl9/aa3324dbd1.jpg
Maybe I should split his mandible into a separate polygroup so I can actually model the inside of the mouth next time, didn't plan for that one. I dungoofed.
LE: are there any issues with posting puush screenshots instead of image hosters ?
I might have made the helmet a tad bit too wide in front, not that many highres pictures of the thing :c
Currently on the chestpiece trying to split up the mesh into proper parts at the suit seams.
I haven't retouched the skin color, need to get back on that once the suit looks a tad bit better.
LE: After this piece is finished rendered in realtime or nonrealtime, depends if I'll be able to make up my freaking mind by the time I reach topology. I'll continue to use this thread as an improvement log and mad ramblings and monologues of late night grammar errors.
LE2: Oh and uh, the suit in original looks something like this.
Suit and meshes are missing any uvs and they have no alpha textures sans the bust. Going forward with the splitting of the suit, where needed.
Maybe I can get some nice people in here and critique this thing so I can hopefully get a couple of fresh perspectives as to why he might look like poop.
That'd be nice.
Screenshots:
So in these suits, what happens if your boot soles would untie themselves in outerspace ?
I'm curious as to what your end goal with this piece is, like if you intend to just finish up and have an awesome sculpt rendered out, or 3d printed, or if your looking to transfer all the detail into a game-ready mesh, or what?
After that I expanded on it with the suit, been slowly chipping on the details trying to make him more lifelike.
Shitty collage with the mercury project suit:
Followed the source pictures and pieces together a suit, the mercury project had different features for each astronaut, so I've taken the liberty to add or remove features that were available to some while being unavailable to others.
It's a good exercise on all levels, organics and hardsurface alike. As to answer your question(which you've hit the nail on the head with), I've been leaning to a non realtime render, vray ss2/photorealism, sort of how ysalex does it, at least to learn the vray shader process for the bust if not for the rest of the suit.
Are there any mesh density/polycount differences between a game mesh and a nongame mesh ?
After I model this guy, I'm going to pose him in one of the Mercury 7 space capsules:
Maybe I should just render the bust in vray and focus on building this whole ensemble as a game mesh ...
Thanks again.
For a portrait-like render, I'd say your nearly (if not already) at a good finishing point for the astronaut, and should look into getting a nice render setup going. Take into consideration what you want to show off the most. If your really just trying to make some cool pictures to wow people at first glance, you could get pretty in-depth with your setups, like by taking multiple renders of the same pose with different lighting setups, and then composting them all together in Photoshop. Work on getting some basics of lighting down too so you're not just rendering with default lights (things like 3-point lighting with a key, fill, and rim light, that sort of stuff). The posing idea you have would be great too (T-poses are pretty boring unless you're showing it for a reason). It takes a bit of extra work on a non-rigged model, where you're pretty much just carefully masking limbs and using the move/rotate tools at different joints to position things around, but if you dig around on that subject, I'm sure there's ways to do it better and easier than that (I believe the transpose master tool is extremely useful for things like that). Also remember that for those kinds of renders, what you don't see doesn't matter, so if you have some messed up geometry or textures in some places from moving him around, it wont matter if the render-cams aren't showing it. If you want to take it a step farther than that, look into setting up a subsurface scattering (SSS) shader network for the skin. It's a pretty complex process to get set up correctly, but it does add a whole lot to your model by making the skin look like skin instead of plastic, and would be a very useful thing to understand if you intend on leaning towards a rendering path in the industry.
As far as turning him into a game character, that starts a whooole new half of production from where you are now. The art of it revolves around taking those millions of polys and multiple hi-res texture and shader networks and simplifying them down to something that a game engine can render and animate in real time, so we're talking in the thousands of polys with maaybe two or three (if not just one) texture/normal/spec/alpha maps for the whole model (all preferably around 2k or less), and that's just kind of a generic estimate. Those are definitely not set in stone, and all those limits vary from game engine to hardware to project/company, even to the importance of the character being made, but its at least a general idea.
If you ask people for specifics for game meshes, usually (I find) that you'll either get too many answers with no point of reference, or too vague of comments on the matter, so instead I'll just tell you about what I experienced.
ArenaNet had an internship test that was to make a character model based on a single concept art picture. You had one month to make that character in high detail (I used a combo of Maya and Mudbox at the time) and then transfer all that detail down onto a low-poly mesh, and they actually specified the limits of the final mesh being 7500 triangles, and only one 1k map each for color, normal, spec, and alphas for the whole character, and the alpha map had to be 1-bit (solid black/white, no grey areas). There's a whooole lot to consider that you don't realize until you start something like that, like figuring out what bits need to be modeled vs. which ones need to just be detail from the normal/alpha maps, how to keep the edge flow clean enough that the character can be rigged properly (like making sure the joints have enough detail to bend correctly and that things wont have a lot of penetration when moving, etc.). In most instances like that, its simply too much of a hassle and too time-consuming to try starting out with one low-poly mesh and then just building more and more subdivisions, because so much can change in your edge-flow when you start adding detail. The most common way of getting the game mesh (as far as I'm aware), is to make the hi-res sculpt first, then build a whole new mesh around that sculpt. After you get the low-res mesh at the polycount you're aiming for and double check your edgeflow and do all your UVing and such, then its just a matter of transferring the detail from the high to the low, which you can do in different ways depending on what program your trying to do it in (in Maya, its the Transfer Maps tool, but most programs have a way of doing it). That has its own kinks to work out, but once you understand what its doing, its fairly straight forward in terms of tweaking settings and cleaning up any problem spots (what it basically does is raycast-render the high-poly detail into color/normal maps for the low-poly). There's a whole lot of little things that you figure out along the way, like getting the high-detail into Maya in the first place to be able to transfer stuff (Maya can't handle millions of polys like Zbrush or Mudbox), how to fix UV seams when you transfer detail into the normal map, how to actually build the low-poly model around the high-poly, etc.
As far as choosing which path you want to do, right now my main suggestion (and remember, I'm by no means a voice of authority on this matter) would be to narrow down which part of the process you enjoy the most, and find out the career path you want to start down on based around that.
If you're also looking to improve your accuracy and understanding of anatomy, you should do more practice pieces. The astronaut is a great portfolio piece, but if you spend too much time on one model, you become too narrow-sighted. Maybe set a goal to create a new sketch-sculpt every weekend or something and just see what you can do within a day, try just doing a highly detailed torso, or a complex pose, or just a bust, etc. And as practically everyone else would say, practice sketching in 2D just doing like traditional drawing too (that's one thing I always kept hearing for animation too, but never had the time or drive to really get in the habit of doing. Spent too much damn time enviro-building in Minecraft instead :P).
As far as actually learning all this complex crap, if you can't find it on YouTube or Google, there's a bunch of tutorial websites that go into specifics. You have to pay for them and sometimes they don't quite cover a specific problem you may have, but they're good starting points at least. I would suggest Digital Tutors to start with, simply because they have such a large selection of tutorials, from specific programs to specific areas of production. You'll undoubtedly have to wade through a bunch of stuff that you already know to get to the crap you really want, but even the basic stuff is still worth the listen, because if you don't pick up something you may not have known, it'll at least reinforce that what you've been doing is correct. And the $30 a month is a pretty nice price, just so long as you really delve in to get at least a few hours of tutorials in per week (or just make sure to cancel the subscription if you're done using it)
Posting the 3d scene as well. https://www.artstation.com/artwork/mercury-astronaut-bust
Moving on...