If you make sure a loop goes around all the hard edges of models, they always naturally have places for loops. That is the true correct way to do it.
I think insetting the sides will work here. Maybe twice is needed, since there are ngons.
That will give you the loop around the edge that you need, and any messyness will be confined to the inner flat part, and so it won't effect your model in a bad way.
You should probably remove those horizontal lines and add back any you need after the inset. Usually you won't need to remove and add, but here you have some topology that will inset over itself.
Things to learn about are loop redirection (changing which way the loop goes so the edge ring goes around hard edges), edge termination (ending an edge loop), and extraordinary vertices (corner vertice, 3 edges to it)
I said inset the sides. He already had support on the front part, and your example seems to be missing needed support anyway, so it's not surprising it came out that way.
Inset is correct here because of topology rules. Like that, with those other parts inset yes, indeed it's wrong, so it's really good you show the example. But an inset just on the sides would put a ring where one should be on hard surfaces, following the corner there, and give all the support he needs, turn that triangle to a quad, and produce a much cleaner end result.
I know your first example works, but it's not necessarily ideal in terms of topology. Not being argumentative, your input is golden as well.
See the triangle on the side of it? With my suggestion, it is all quads (except perhaps for planar ngons).
In other words, I can make that shape in all quads where quads matter with the inset.
I need to run, do you have time to make one in the way I explained? I can throw one together when I get back if it's not clear. If you can produce one as I explain, it'll mean I explained it good enough, and save me from showing one :P
Alright thanks guys. I think I got it, I'm still learning. I just added inserted vertices and connected them to form the edges that I needed. Probably not the fastest way but it worked. = )
I've been using subdivision edge creasing on top of the initial support edges a lot lately, just because it allows me to work with less fiddly geometry to get the results I'm after....
If you plan on taking this into ZBrush or Mudbox you'll want to make sure your poly distribution is more even, for sculpting. The wireframe shows it's pretty sparse in some areas and extremely tight around the creases. You should add some loops around the width and length to even that out some (if you plan to sculpt).
This reminds me, someone should make a script which takes height and width of faces and applies set of connections to it, to even the flow and distribution, for much nicer ZBrush sculpting.
It seems like you don't have a core understanding of what way support loops are used and when not used. Also, you can yourself alot of paint if your model floating geometry. Hell, from the looks of it, you can just detach a plane, model it, and put it ontop of your hole, but without the extra need for weird polies.
Here are some tutorials on to get a hard surface modeling going on:
Yeah, sorry. I really need to relearn all this stuff. I've watched videos in teh past but haven't really applied the stuff to any real projects until now. Sorry I'll start researchin'. = (
With all due respect EarthQuake, that thread is fucking ace and bloody awesome, and it has saved me alot of time, but it's pretty cumbersome to move through the thread at certain places and it can get pretty daunting for people who aren't into this area of hard edge modeling.
For example, when I first read that thread, and you guys said "throw more polies", I ended up doing just that and wasn't too happy with it since it more times then not killed my mesh, so I then started "throwing more polies" in conjunction with "make it equally distributed", even my flat models now look more natural since I'm not simply limiting myself to support loops but an overall flow which retains my model more.
I still keep the thread next to me as a guide, but for some people, it's pretty daunting, and info that should go hand in hand aren't near each other to allow a quick paced understanding.
I'm sorry Pior if that was the impression I gave, I can say I have read everything in that thread atleast 3 or 4 times, and know which pages to reference at my lab for a quick refresh. I even have segments of that threads saved out as a PDF.
So no, too much info isn't the problem, but I can say this, 2 of colleagues where scared to look at that thread and I had to push to one of them in asking a question in that thread. They felt they would look stupid if they asked a question that 'may' have been answered or that they already saw but didn't really grasp it because some other post said something abit different.
I think that looking at your models the key thing you should be trying to do is to keep your control cages is quads. This will stop you from causing the problems you are having. The Ngons are where you're getting strange smoothing.
Like Pior said - take the time to read that thread. It's huge, yes, but it's invaluable.
Although keeping quads is very important, there are times were Ngons are perfectly fine.
The main problem you have in that example scizz, is that your edge loops are incomplete. You were doing just fine, but it looks like you stopped before completing them.
The red loops are really just continuations of what you already started, then stopped. There's an additional red loop on the outside of the left and right sides, in order to give support to the inner loops and give a nice clean edge. With the red loops alone, it will get rid of all of the wonkiness you have.
The green loops are support loops for the side walls, making them have a crisper edge. Without the green loops, the indent will be curved like an inset cylinder. Those green loops will make it nice and square.
They felt they would look stupid if they asked a question that 'may' have been answered or that they already saw but didn't really grasp it because some other post said something abit different.
To fear looking stupid is to ensure you will remain stupid.
I understand the feeling (it definitely kneecapped me in the beginning) but there is so much truth to the statement "the only stupid question is the one unasked."
Replies
I think insetting the sides will work here. Maybe twice is needed, since there are ngons.
That will give you the loop around the edge that you need, and any messyness will be confined to the inner flat part, and so it won't effect your model in a bad way.
You should probably remove those horizontal lines and add back any you need after the inset. Usually you won't need to remove and add, but here you have some topology that will inset over itself.
Things to learn about are loop redirection (changing which way the loop goes so the edge ring goes around hard edges), edge termination (ending an edge loop), and extraordinary vertices (corner vertice, 3 edges to it)
Not the best but you get the idea
Edit:
@mortalhuman: Please don´t use inset it almost every time fails to get the result you really want
As you can see here:
Inset is correct here because of topology rules. Like that, with those other parts inset yes, indeed it's wrong, so it's really good you show the example. But an inset just on the sides would put a ring where one should be on hard surfaces, following the corner there, and give all the support he needs, turn that triangle to a quad, and produce a much cleaner end result.
I know your first example works, but it's not necessarily ideal in terms of topology. Not being argumentative, your input is golden as well.
See the triangle on the side of it? With my suggestion, it is all quads (except perhaps for planar ngons).
In other words, I can make that shape in all quads where quads matter with the inset.
I need to run, do you have time to make one in the way I explained? I can throw one together when I get back if it's not clear. If you can produce one as I explain, it'll mean I explained it good enough, and save me from showing one :P
And if he did it the "right" way he should be using inset for anything(supporting edges related)
Wink, wink...
EDIT: Did you watch this? http://cg.tutsplus.com/tutorials/autodesk-3d-studio-max/3d_cg_vfx_modeling_autodesk_3dsmax_subdivision_smoothing/
It seems like you don't have a core understanding of what way support loops are used and when not used. Also, you can yourself alot of paint if your model floating geometry. Hell, from the looks of it, you can just detach a plane, model it, and put it ontop of your hole, but without the extra need for weird polies.
Here are some tutorials on to get a hard surface modeling going on:
Tire: http://eat3d.com/free/modeling-tire-3ds-max
Support Loops: http://cg.tutsplus.com/tutorials/aut...ion_smoothing/
A bunch of stuff you might wanna look at in spare time: http://vimeo.com/user3618217/videos
Again, another video on idea of HS modeling: http://vimeo.com/10941211
For example, when I first read that thread, and you guys said "throw more polies", I ended up doing just that and wasn't too happy with it since it more times then not killed my mesh, so I then started "throwing more polies" in conjunction with "make it equally distributed", even my flat models now look more natural since I'm not simply limiting myself to support loops but an overall flow which retains my model more.
I still keep the thread next to me as a guide, but for some people, it's pretty daunting, and info that should go hand in hand aren't near each other to allow a quick paced understanding.
So no, too much info isn't the problem, but I can say this, 2 of colleagues where scared to look at that thread and I had to push to one of them in asking a question in that thread. They felt they would look stupid if they asked a question that 'may' have been answered or that they already saw but didn't really grasp it because some other post said something abit different.
Although keeping quads is very important, there are times were Ngons are perfectly fine.
The main problem you have in that example scizz, is that your edge loops are incomplete. You were doing just fine, but it looks like you stopped before completing them.
The red loops are really just continuations of what you already started, then stopped. There's an additional red loop on the outside of the left and right sides, in order to give support to the inner loops and give a nice clean edge. With the red loops alone, it will get rid of all of the wonkiness you have.
The green loops are support loops for the side walls, making them have a crisper edge. Without the green loops, the indent will be curved like an inset cylinder. Those green loops will make it nice and square.
To fear looking stupid is to ensure you will remain stupid.
I understand the feeling (it definitely kneecapped me in the beginning) but there is so much truth to the statement "the only stupid question is the one unasked."