[WIP] Character - Gina (Critique wanted)
triangle
After browsing this section for a while I've come to a conclusion that I'm supposed to start to post only if my project has something substantial to show. But then what if I commit to a grave mistake early on? I do need help with the whole thing after all. Deciding to just go for it, I'm starting this WiP thread, sorry X)
This is a pretty straightforward interpretation of Unreal's default character - Gina.
I'm aiming for something relatively simple in design, down-to-earth materials. I do want to play up some of the original model's... perceived features.
One such feature is a "hooked nose", been dead set on that one for years =D
The goal is to push for a more realistic model - something that I don't normally do, as I don't have an eye for realism. I can enjoy it, but how to achieve it?..
Started on the body, it's still easy enough to change things around.
Goal: a somewhat tough-looking lady, but not "super-human super-model" type. Can use just a little more volumes all around I guess? Or not.
Already not sure if proportions are realistic or not. Especially with the arms and hands. Are they tiny?
The head so far. Never made a fully realistic one before, so quite a bit lost here.
Any help would be most appreciated! Thank you for your time.
This is a pretty straightforward interpretation of Unreal's default character - Gina.
I'm aiming for something relatively simple in design, down-to-earth materials. I do want to play up some of the original model's... perceived features.
One such feature is a "hooked nose", been dead set on that one for years =D
The goal is to push for a more realistic model - something that I don't normally do, as I don't have an eye for realism. I can enjoy it, but how to achieve it?..
Started on the body, it's still easy enough to change things around.
Goal: a somewhat tough-looking lady, but not "super-human super-model" type. Can use just a little more volumes all around I guess? Or not.
Already not sure if proportions are realistic or not. Especially with the arms and hands. Are they tiny?
The head so far. Never made a fully realistic one before, so quite a bit lost here.
Any help would be most appreciated! Thank you for your time.
Replies
Remade the forms. Abs seem less broken now. Still have doubts about the arms.
Now I'm not sure how to proceed further D= The face seems off-target and I can't see why.
I was making cream for a cake today, and it feels like a very similar process: "Oh no, it's still too liquid. Did I mess up the ingredients? Maybe it needs more time? But it's burning a little. Did I miss a step? There's like only two of those though, how could it possibly go wrong? Well darn it, it's never going to work, the blasted thing is ruined now, it... oh wait, it's done"
The point is, I guess, there's no need to be so dramatic every step of the way XD
The head and hands of my current basemesh are separate from the main body because it wasn't easy to resolve different loopcounts. I just shrinkwrap a few loops on edges between these objects a couple of times - haven't failed me so far, but I guess it only works, because there are no small details in overlapping areas.
Yeah, I always check proportions this way (since the hand is easily detachable =D). But I always doubt what I see anyway =')
Here's what I've got:
____
It occurs to me, that I should've put a firm time limit. It's not like I'm in a hurry, but there are other things I want to try sometime in this lifetime.
With this in mind, I'm wrapping up "blockout" stage. I thought about adding armour, but I'll leave it for later. There enough new things here to learn as it is.
First time trying out Blender's Cloth brushes. It took a while, and I still don't quite understand how to control the results.
The resolution of the mesh influences the size of the folds - can't get small folds if there's no geo there, but at the same time, it seems hi res geo can only result in small ones.
Drawing XYZ-folds beforehand seems to almost guarantee that these will get reinforced by the simulation though.
I do love using floating geo for seams. So crisp. So far had no issues baking them, though my previous tests were more stylized.
But it only feeds my indecisiveness (and slows down performance eventually), so maybe I should turn these to meshes sooner rather than later ='D
_____
Meanwhile:
I have no idea how I'm going to retopo and bake this harness thing and I'm definitely going to regret those loops on pants if I leave them like that...
Probably have to rethink the gloves, because they still don't match the rest of the clothing and it looks odd.
Other than that what's left to do is another iteration of the face, collapsing the last of non-destructive geo, adding wrinkles/imperfections here and there on the cloth and... I feel like I'm supposed to be doing something more, but I don't see what that could be.
Seems like the "trick" was to use very low-pressure Draw brush to gradually build it up. It's easy to zone out and completely forget about structural considerations though. Having PureRef open during the whole process helps staying focused a little bit, or at least reminds me to kick myself to attention once in a while and make conscious decisions about what to put where.
Not sure which tutorials I should focus on though. I'm learning a lot just reading through this forum and wiki, but it's chaotic as I don't really know what exactly I'm lacking.
____
Well, at any rate, I think this is as far as I can push the hipoly at this time. Maybe add slight face asymmetry, but that's it.
Time for retopo then D= Except...
...how straps and belts are normally retopoed in these cases? Or rather, how the empty spaces between them are handled?
Feels like in this case I'm supposed to glue the belt loops to the belt and retopo it as one solid geo... but what if I want these additional loops below?
Just retopo the whole loop as a separate object? That would be much easier to unwrap, but harder to pose/animate...
As for the belt loops, smushing them down and maybe filling them up might be the most optimized solution, depending on how close you want to get later, but keeping them separate should be fine, too, if you match up vertex positions with the main mesh (basically cut the intersections of the loops into the main mesh) so that there's no unexpected interpenetration during animation.
Hmmm, to be honest, I'm a little reluctant to touch the anatomy at this point, though I can probably smooth the transition from deltoids a bit.
Now it's the question of getting the resolution right. I guess, I've added too much, going to even it out a bit. The point is for it to look good enough at "portrait distance". And to go with haircards (which, however I look at it, will take A LOT of polycount).
But this is the moment where I make things weird. Here's my reasoning - please, tell me if I'm making it more complicated than it needs to be:
Now I know that it's normally recommended to keep things in quads. Of course, a "game" topology is tris anyway, but quads still play better with editing tools.
But conforming to ridges and cavities (like clavicles) AND keeping it all quads bumps up the polycount pretty quickly (yes, you probably can't tell I care about such things from my face topo =D ). Plus there's "animation flow" to consider.
So. If one were to start aggressively triangulating parts of the mesh to ensure a "secondary flow" of sorts... is it legal? For mid/high res models, I mean (I imagine it's a must for low-poly). Or are they supposed to be as "pure" as possible (with possible exception of areas of resolution increase/reduction).
Is some kind of topology purity police already coming after me? D=
This one is pretty much the same as the "sculpt" topology, but cut and adjusted along pronounced plane changes:
Neither of these meshes are final, of course, plus I'm still figuring out the density.
I'm convinced that my current retopo is a failure and it would be better to redo from scratch now, then keep trying to "fix" it.
Should've started with higher density and worked from there. It seems harder to increase, than the other way around.
In this particular work, it's probably better to do it more detailed.
After reading "waviness"-thread (pure gold, by the way, thanks for keeping this forum and its knowledge alive
I'm looking through some "modern" game examples right now, and those models I could find with wire look like 4 times more dense than what I'm doing... The level of details I'm looking for is something like Devil May Cry V or Jedi: Fallen Order.
I think I only ever "finished" retopo when I was tired of guessing the right way to do it and gave up on potential problems
We are probably looking roughly along the projection direction, here, which is why the lower/outer edge of the rim looks smooth. If the top wasn't a seperate object, the transition to the skin would likely show the same smooth curve (from this angle or a bit higher, which is close to eye height, so not a bad compromise, and assuming it isn't hidden by the geometry).
Since it is a separate object, we don't have that advantage, but any sloping etc. won't help much either, since this is simply your geometry showing. It's not that the edge looks awful, but that the lower/outer edge of the rim looks much smoother in comparison. A sloped rim (which you could do in the LP only, too) might help you to even that out from different viewing directions or shift the projected edge away from the geometry edge so that the angle change doesn't alternate across the latter and/or we see both sides of the edge from a wider angle (less chance to see the silhouette), but it can't change your geometry, which might have added to your decision to aim for a higher polycount.
In this case, because we don't have to see the inside of the top, you could indeed model the rim with a slightly flatter angle than 90 degrees from the start and it might improve things a bit without detracting from the realism in a meaningful way. But you are also correct that this doesn't apply for shapes that are equally visible from all sides and since you can't make it too obvious, it's probably not worth the trouble at this point and not as impactful as increasing the resolution.
Like you have probably gathered from the wavyness thread, there will almost always be some warping from some viewing angle if the geometry is different enough. If you want to avoid that completely and have a dense and regular enough mesh for the overall shape, you could e.g. do your normal details in 2D directly for your lowpoly or bake from a subdivided lowpoly (assuming high and low are more or less the same SubD mesh with the same topology and the only difference are details like ridges or seams). In this case, that would mean that all elements would be as angular as your outer edge, though, just consistently.
As for much higher LP resolutions, I don't have any experience with that, but I'd wager at one point it makes sense to stick to quads and subdivide. Or at least use tools like quad fill and do triangles at the very end (if necessary at all, which seems less likely with increasing resolution). The basemesh for high and "low"poly could even be the same.
For what it's worth, I wouldn't increase the resolution at this point and finish with the old specs. This is a bit like someone throwing a spanner in the works from outside, only you are doing it to yourself. But it's your decision.
After some agonizing consideration, I've decided to go the high(er)-poly route after all (among other things because it doesn't really cost me anything other than the time already spent and a little self-esteem - I didn't decide on any definitive poly limit at the start after all).
Beside looking up some examples, I went over my own models and figured I never actually tried this approach. It's like I panic every single time when polycount rises for some reason XD
The problem here it being often at odds with what I'd like to try: smoother shapes, higher res haircards. Maybe doing things differently will be beneficial. Could get a better feel for different geometry density and what it entails. Maybe. Hopefully.
Stop running around screaming, do things hipoly, use enough texture res, focus on the face and eyes, and work on better hair - guess that's the plan.
None of this feels right, but I have to stop somewhere.
Around 85k tris. Plus I estimate a few dozens k for hair, so I guess it's fine, but man, what am I even doing...
ok then <inhales> AAAAAAAAAAAAAAAAAAaaaaaaaa........
After packing and re-packing UVs, then doing test bakes, then changing my mind and re-packing again, I've ended up with 4 x 2k textures sets...
Everything's baked fine (once I've fixed all the spelling mistakes), as far as I can tell. Set up layers and filled in some base properties... now what?
But yeah clothing textures isn't too tough, but it can be time consuming!
____
Now that I made a screenshot, the face looks weird. Not sure if it's because of lack of eyelashes, or if it really is strange.
Still need to adjust colors and paint skin roughness map.
Time to assemble final materials in-engine?