Vehicle : Hearse
polycounter lvl 18
Having a go at my first vehicle and need some support to see me through it. I have some good reference and started to block it out. Main problems I think will be...
- Tricount. From what I have so far, I think the final LP tri count will be around 7000. I usually like to model really low poly but am trying to push myself here to something that's even possibly beyond 'current gen'. I also need to ask myself if I should build an interior if the count is already unusually high. Will be using zbrush for deco and high poly baking for a few elements.
- The parts connecting the wheels is a bit daunting at this stage. Defiantly going to have to find a way of simplifying that. At the moment I can't make no sense of what exactly is going on there. I'm not too fussed about perfectly copying the reference.
Reference + WIP
- Tricount. From what I have so far, I think the final LP tri count will be around 7000. I usually like to model really low poly but am trying to push myself here to something that's even possibly beyond 'current gen'. I also need to ask myself if I should build an interior if the count is already unusually high. Will be using zbrush for deco and high poly baking for a few elements.
- The parts connecting the wheels is a bit daunting at this stage. Defiantly going to have to find a way of simplifying that. At the moment I can't make no sense of what exactly is going on there. I'm not too fussed about perfectly copying the reference.
Reference + WIP
Replies
I wouldn't worry about reducing the count any further - why not do the model justice, then worry about optimizing? 7k doesn't even seem that excessive to me, and if you put too much thought into reducing the count as opposed to capturing the essence of the vehicle, the end result could suffer. I say get it nailed down, then start trimming.
I can see that the front wheel connection could be a little confusing, but the image looks like it was taken in some kind of transport museum - can you take any more reference? Or find out the specifics of that particular carriage, and get some alternative angles? If not, I guess some artistic license will have to do.
Keep going!
update, working on the fiddly bits. Think I've simplified yet kept it looking functional. From what I see in my ref pics a lot of it is added bits meant for keeping it still because it's a museum piece.
Not sure if to start adding extra elements now (Some carts have bits on top, steps, flowers etc ) or to go in and clean up and get ready for zbrush/high poly baking etc.
I'd get the steps and other features modeled in now before exporting to ZBrush. Doesn't look like you've got that much to do before going in and sculpting all that juicy detail on the panels and such.
Question, all my cylinders for bars etc are mostly 8-12 sided. Is it really worth making highpoly versions of these for baking? One thing that puts me off is that it means each bar needs it's own unique UV space where as usually I'd just making a tiling bar texture and map them all on that. I have no strict budgets here but I want it to be somewhat realistic in terms of budget.
Can't wait to see what you will do with it..
Something tells me Danshewan is right and I should just forget it, make it look awesome and try to optimize as much as possible afterwards but I honestly still see it being 10-15k. It's now 8k...
The whole point of this is really to show I can do modeling, normal baking and zbrushing. Even if it ended up being 15K it still shows those skills right?
As for the normal map issues, chamfering edges is definitely one way to do it, but there are other ways also, like using multiple smoothing groups, and separating uv chunks.
I found this article to be extremely helpful with understanding all the tricks/ in's and out's of tangent space normal maps: http://www.svartberg.com/tutorials/article_normalmaps/normalmaps.html
By the way, I had a peek at your portfolio, and you have some really awesome stuff in there. So just take your time with this, and make it as awesome as I know you can! Detail it up with high poly
Started normal baking today with part of the seat. So as it stands this would be the final piece minus the zbrush work, my plan is to only do deco stuff in zbrush on flat planes then copy the normals and paste into my psd.
There are a few little artifacts here that I can hopefully fix in photoshop though I quite like some of these artifacts, some could be read as kinks/chips in the wood etc.
I just mirrored it all and remembered that doing this screws up the normals! Think the only way around that is give each element that requires normal baking it's own unique UV's and rebake? there must be a solution because I'm pretty sure most game assets even today have many mirrored elements?
I thought at one point that I had tangent space normals completly understood, then I ran into problems on some hard surface objects (a gun) and realized that they can be much more difficult/complicated than I thought.
One thing you can consider is using object space normal maps to avoid any problems like you are having (Marmoset supports them as well as other emerging game engines). However, I believe that because tangent space normals are pretty much the standard that learning all the in's and out's is better and more important / impressive.
Anyways, there are two things that I have noticed that commonly will cause seams in normals when you are mirroring. (the seam will be at the mirrored edge) This is assuming you have a good and correct bake.
The first is smoothing group issue. If you have different smoothing groups along the mirrored part, you will have a seam.
The second is due to angular change.
I wrote up a big article but decided the best way for you to understand what I mean is to simply run some tests of yourself. (I'm not sure what I was writing actually made any sense lol).
Anyways make a simple cylinder in the top view. Bevel from the top and bottom a few times and make it so you have flat edges along the z axis and angular edges along the z axis.
Make a fancy high poly, turbo smoothed column to bake off of that fits in your low poly as you normally would.
make your uvs and cut the cylinder in half along the y axis, so that you are attempting to mirror over the y axis.
Once you bake show your normals with a real time shader, you will notice a nasty seam along the y axis.
But if you render the object, the seam disappears!
This is due to how the shader renders realtime vs built-in scanline which is more expensive.
If you made your low poly cylinder like I said, and you have flat polys going straight up the z axis in some areas. Select those faces and detach them. You will now see that the seam is gone alone this part of the low poly!
This is because you no longer have the angular change affecting the normals as you did before in these areas. Play with this some more and I'm sure you will get what I mean about angular change.
Anyways, in your case, if you mirror along what appears to be the y axis on the hearse, and you have your uvs mirrored as well, and you have the same smoothing group setup for those polys across the y axis; you shouldn't have a problem with seams because there is no angular change in this area.
A good way to simply test if the problem is the angular issue mentioned above is to render out your low poly object, and if the seam disappears then you have to work on your setup by detaching stuff setting up separate smoothing groups or other options.
But if its a baking issue, it could be something as simple as you not moving mirrored uv islands during the baking process. Or attempting to bake flipped uv islands, and stuff like this.
I hope this helps, and if you could give me a specific example of where your problem area lies, perhaps I could figure out what is going wrong. Are you using 3ds max scaneline renderer to bake your maps?
high poly > low poly > low poly with normal bake
Thanks Brad for the large info surge, will read it properly tonight. You've helped me in other threads too, really appreciate it!
I see that you gave these areas their own smoothing group on the low poly which is correct, however, if you have the uv's attached in the areas where there is a change in the smoothing groups it will cause a seam.
Like I said, I'm not sure if there is even a seam in these areas, but it looks like there might be.
Glad to help buddy!
Also not sure if you're misreading my mirror problem. There isn't a problem with seams but incorrectly flipped RGB channels. My mirroring mainly will be along the x axis, for example mirroring one wheel from one of the sides to the other.
Ah interesting. Yeah the cone on the top has a seam, the uv's for this our connected. The top edge of the Umbrella has NO seam at all and the uv's are NOT connected. Similar to what that document you showed me was explaining. So what I do now is split the cone uv's, rebake and it should be fine!
edit : yep, fixed and updated previous image showing the lights
Good point. I do turbo smooth them but maybe they are still low. For example if I'm doing a cylinder I usually just make em 24-32 sided, maybe I should just make them 100+ sided to make them perfect or something but I'm working on quite a low end laptop. As it is I'm having to split this all into multiple max files to make sure my laptop can handle it!
Done all but the micronormals and deco now. I think I'll be better off doing the small deco (like on top of the pillars) in photoshop/nvidia the rest I'll try and do in zbrush.
Does anyone think the roof needs a few more subdivision to round it off more? So instead of a triangle it's more like /=====\ (with ==== on top). Does that make more sense?
It's currently 13.5k with a 2048 map.
Adding deco now...
maybe the folds in the cloth could use a bit of tightening up? Have you tried using the damstandard brush? or one of the brushes from here http://www.selwy.com/2009/zbrush-clothes-tutorial/
The wheels jump out at me now, noticably lower poly than the rest. If you look at the source at the top, the profile is much different. The spokes don't come flush to the edge of the rim, there's a bit of a bevel on that. They're also more flat than squared off like that.
OrganizedChaos Thanks for the tip, redid the cloth in zbrush with those new brushes
throttlekitty Thanks, adjusted the spokes.
now texture that sucker, itll be an impressive portfolio piece. You inspire me to start something similar, heh.
Can I get away with a 2D texture on those lamps? or do I really need to remodel them and make an interior?
Bit of a lame screenshot but just a quick WIP. Not really started on the wood yet but starting to add colour shifts and edgework to fabrics and metal. You'll also noticed I've changed the lamps so they now have an interior.