Finally finished my first car. The most complex model I've made before that one was a sofa so I consider this to be my first "serious" model. I would use totally different techniques if I started it from scratch but what's done is done. There are quite a few flaws which could be fixed pretty quick but I may some obvious ones SO CRITIQUE IS WELCOME I'll continue to model interior and some other details but this is the base model which I'll use to make low poly.
So is there anything I should fix before I make low poly? How is topology in general?
I'd agree with Zac... it looks like it'd fall off at higher speeds or get torn to pieces by ice. Also... I came here expecting an average snow mobile... I am not disappointed sir, you have bested my expectations. good on you.
otherwise, aside from the "cage"... it looks fine right now.
I made the cage more bulky. I'm almost done with low poly which is pretty heavy right now but I intend to bring it down a bit more. Now I'm unwrapping the whole thing.
That's not the lowpoly right? I see plently of edges doing nothing!
Also, that small cylinder at the back seems to have more sides than the big engine turbine.
That's not the lowpoly right? I see plently of edges doing nothing!
Also, that small cylinder at the back seems to have more sides than the big engine turbine.
it's not a final low poly but since it's my first one can you point out which edges I can remove safely for example? Am I doing this right:
no you re not doing it right, you re just creating n-gons (polygons with more sides than 4) ...the resulting triangle count will be the same, as the engine will triangulate the polygons in the end.
the ski-thingies look out of style than the rest of the vehicle. you should look into some period correct hydroplanes to get a similar style into those. you could add some wire mesh on the sides of the propeller cage for extra protection.
i suggest you retopo the whole thing over your high poly, don't use your un-divided lowpoly as a realtime mesh.
I tried to retopo a part of the body and results were subpar. I'll explain further. I also found a double mesh in the process. deleted it and cut down 10k of triangles so it was worth it in the end.
So, let's take a look at body's triangle count:
It's only 10k, which is quite reasonable I think. And it's not what I used for high poly, I deleted a lot of supporting loops.
Meanwhile, the transmission - the part which almost nobody can see sits at 10k.
Chrome parts sit at 10k too.
And the interior is only 2500 tris.
Meanwhile retopoing this gave me 3 times less tris on the body but it looked horrible and about 30% less tris if I added more loops to preserve curvature. I think it's more important to cut optimize parts people don't pay attention to like transmission than try to save polys on the part which is in focus. I plan to delete most of transmission and redo some chrome parts. It will bring polycount down to 30k (now sits at 46) which is reasonable.
you ve spent too much polygons around the front fenders because you are trying to keep them unified with the body mesh. you dont need to do that. in reality, they are separate parts, so you dont need to take so much effort to make them match vertex for vertex. make the fenders as separate elements, and use how many polygons you need.
as for the suspension, the springs are what takes all the polygons. simplify them, or even make them a cylinder and use an opacity map to fake the spring. or use your imagination, remove the spring and give it air-ride suspension (bags)
you ve spent too much polygons around the front fenders because you are trying to keep them unified with the body mesh. you dont need to do that. in reality, they are separate parts, so you dont need to take so much effort to make them match vertex for vertex. make the fenders as separate elements, and use how many polygons you need.
as for the suspension, the springs are what takes all the polygons. simplify them, or even make them a cylinder and use an opacity map to fake the spring. or use your imagination, remove the spring and give it air-ride suspension (bags)
thanks, Ada. I never thought about using opacity map, I'll look into this. As for fenders - I haven't tried to keep them unified with the main mesh. They take so much polys because there are many bevelled edges and they need support loops to make them sharp. The irony is that you can't even see them when you do a normal map because when I was bevelling edges I didn't think of how they would look in normal map. Guess what, they suck because they are too thin for such resolution. Reflections suck too by the way and I don't know how to fix them except bumping up resolution.
I need to redo a lot of parts to make them look good in normal maps and I need to retopo the main body of the car because it doesn't look smooth with normal maps. I want it to be glossy and when it's glossy it looks pretty bad.
Also I tried to add bolts and metal plates and failed miserably. The best software for this is nDo or some sculpting software and I haven't looked into this at all yet. Doing this in Substance Designer wasn't successful becase I need to warp texture in some places and you can't do that in SD yet. As for Substance Painter you can't draw straight lines there yet and I need them for my bolts.
So the plan is to go to vacation next week, return and dive into sculpting + texture painting.
I did this model to expand on my basic hard surface skills and I learned a lot but I also made a mistake thinking that every high poly would suit for a game model after making low poly and all that stuff.
Also after I fix everything I will add guns! Next step is to rig it, take it into UE4 and make it drivable. This will be my go-to model to learn basics of everything until the next spring I think.
i dont know if you re planning on making it damaged , metal plated etc,
but you should know, all racing games don't use normal maps on the cars , at least not on the body panels, it's a waste, it will mess with reflections, and its cheaper to make it with polygons, rather than huge textures.
Aga22, shit, I've often thought it sounded kinda ridiculous that games like GT have something like 100k polys per car, but that makes so much sense now. Thanks for enlightening me!
@Aga22: ah, it's so liberating to know that it's not my bake sucks but the general approach is wrong. I think I'll increase polycount on body panels and all reflective surfaces substantially. The weapon will be low poly though. And yes, I planned on adding metal plates on skis and the engine. I also wanted to make 3 versions of the car - a brand new car, an old car and a damaged car. If I want to add only rust in the normal app should I also bake body's geo as well? I can't try it right now unfortunately.
After a year of learning I decided to try making the low poly version of this again and this time everything clicked. No simulator specs, just a plain vehicle. I hope to be done with baking tommorow.
Not all parts are in yet, I need to get interior baked and some minor outer parts. I also lost a part of the cage which supported the engine and I think of getting it back but it will need redoing a part of high/low poly, UVs, rebaking, so may be I won't do it. Some modifications of this car used supports like these, it wasn't meant to be fast anyway.
@DireWolf: Modo is good but nothing is perfect. Sometimes I look at Blender (and Max) with their stack and envy their users. But I like where Modo is going. I'm doing this on 901 trial and baking with Farfarer's toolkit has been a breeze. The only part of it I REALLY hate is exploding the mesh. I've tried baking it in Painter with their new "by mesh" baking system but it doesn't support the cage yet and some parts of the model required it. So baking is still PITA. I need to resolve it in my future projects.
The seems/grooves on the floats/feet are a bit large and spaced unevenly they look like you've just put them there to fill up space. I would opt for the less is more approach to them.
The rear wheel tub needs to be cut out as if a wheel could go there, then they have put a panel in to cover it up - its a small detail but its nice, as it has some dimpled lines through it for strength, that add to the design.
Also try and make the fan shroud a bit more streamlined to the body like in your ref, it will make it flow better as a whole piece as the car body is quite streamlined too
keep at it, just look over the ref and be patient it will get there !
@MephistonX: I agree about the wheel tub, there's a panel to cover up in high poly, but I constantly forget to add it to low poly. I also agree about remaking the shroud (and adding supports for the engine while I'm at it). Not thrilled about redoing low and rebaking but I don't think it'll be much work in that case.
Textured a bit more. The model still lacks something but it's getting there. Leaks are too ugly at the moment.
I've decided to redo the tube and pack UVs TIGHT to avoid repeating details on both sides of the car (it was mirrored). Now I can paint damage, bullet holes etc freely. I've managed to do it with the same texel density, so it shows that my previous packing was pretty lazy. Now to run it through IPackThat demo...
Replies
otherwise, aside from the "cage"... it looks fine right now.
Also, that small cylinder at the back seems to have more sides than the big engine turbine.
it's not a final low poly but since it's my first one can you point out which edges I can remove safely for example? Am I doing this right:
the ski-thingies look out of style than the rest of the vehicle. you should look into some period correct hydroplanes to get a similar style into those. you could add some wire mesh on the sides of the propeller cage for extra protection.
i suggest you retopo the whole thing over your high poly, don't use your un-divided lowpoly as a realtime mesh.
As for skies - I've thought about fixing them to match the original reference and add finer details through a normal map. But they are fairly close:
And the cage is pretty close too:
Some models didn't have it at all:
I'm thinking about adding this shape which may be a fuel tank which they started installing on later models:
But I like how car's back looks like and this thing behind would spoil its curves.
So, let's take a look at body's triangle count:
It's only 10k, which is quite reasonable I think. And it's not what I used for high poly, I deleted a lot of supporting loops.
Meanwhile, the transmission - the part which almost nobody can see sits at 10k.
Chrome parts sit at 10k too.
And the interior is only 2500 tris.
Meanwhile retopoing this gave me 3 times less tris on the body but it looked horrible and about 30% less tris if I added more loops to preserve curvature. I think it's more important to cut optimize parts people don't pay attention to like transmission than try to save polys on the part which is in focus. I plan to delete most of transmission and redo some chrome parts. It will bring polycount down to 30k (now sits at 46) which is reasonable.
as for the suspension, the springs are what takes all the polygons. simplify them, or even make them a cylinder and use an opacity map to fake the spring. or use your imagination, remove the spring and give it air-ride suspension (bags)
thanks, Ada. I never thought about using opacity map, I'll look into this. As for fenders - I haven't tried to keep them unified with the main mesh. They take so much polys because there are many bevelled edges and they need support loops to make them sharp. The irony is that you can't even see them when you do a normal map because when I was bevelling edges I didn't think of how they would look in normal map. Guess what, they suck because they are too thin for such resolution. Reflections suck too by the way and I don't know how to fix them except bumping up resolution.
I need to redo a lot of parts to make them look good in normal maps and I need to retopo the main body of the car because it doesn't look smooth with normal maps. I want it to be glossy and when it's glossy it looks pretty bad.
Also I tried to add bolts and metal plates and failed miserably. The best software for this is nDo or some sculpting software and I haven't looked into this at all yet. Doing this in Substance Designer wasn't successful becase I need to warp texture in some places and you can't do that in SD yet. As for Substance Painter you can't draw straight lines there yet and I need them for my bolts.
So the plan is to go to vacation next week, return and dive into sculpting + texture painting.
I did this model to expand on my basic hard surface skills and I learned a lot but I also made a mistake thinking that every high poly would suit for a game model after making low poly and all that stuff.
Also after I fix everything I will add guns! Next step is to rig it, take it into UE4 and make it drivable. This will be my go-to model to learn basics of everything until the next spring I think.
but you should know, all racing games don't use normal maps on the cars , at least not on the body panels, it's a waste, it will mess with reflections, and its cheaper to make it with polygons, rather than huge textures.
Aga22, shit, I've often thought it sounded kinda ridiculous that games like GT have something like 100k polys per car, but that makes so much sense now. Thanks for enlightening me!
@Aga22: ah, it's so liberating to know that it's not my bake sucks but the general approach is wrong. I think I'll increase polycount on body panels and all reflective surfaces substantially. The weapon will be low poly though. And yes, I planned on adding metal plates on skis and the engine. I also wanted to make 3 versions of the car - a brand new car, an old car and a damaged car. If I want to add only rust in the normal app should I also bake body's geo as well? I can't try it right now unfortunately.
After a year of learning I decided to try making the low poly version of this again and this time everything clicked. No simulator specs, just a plain vehicle. I hope to be done with baking tommorow.
Keep going, want to see more
Not all parts are in yet, I need to get interior baked and some minor outer parts. I also lost a part of the cage which supported the engine and I think of getting it back but it will need redoing a part of high/low poly, UVs, rebaking, so may be I won't do it. Some modifications of this car used supports like these, it wasn't meant to be fast anyway.
@DireWolf: Modo is good but nothing is perfect. Sometimes I look at Blender (and Max) with their stack and envy their users. But I like where Modo is going. I'm doing this on 901 trial and baking with Farfarer's toolkit has been a breeze. The only part of it I REALLY hate is exploding the mesh. I've tried baking it in Painter with their new "by mesh" baking system but it doesn't support the cage yet and some parts of the model required it. So baking is still PITA. I need to resolve it in my future projects.
The rear wheel tub needs to be cut out as if a wheel could go there, then they have put a panel in to cover it up - its a small detail but its nice, as it has some dimpled lines through it for strength, that add to the design.
Also try and make the fan shroud a bit more streamlined to the body like in your ref, it will make it flow better as a whole piece as the car body is quite streamlined too
keep at it, just look over the ref and be patient it will get there !
Textured a bit more. The model still lacks something but it's getting there. Leaks are too ugly at the moment.
also like the new color :thumbup:
Rendered it in Unity 5.3
model
Made more accurate color for soviet vehicles, also toned down bolts.
the whole set of images