SOLVED How can i add bumps and better cambers to my track on 3ds max.

How many polys would you say is too much on the low poly version if I add 1 iteration of mesh smoothing to each corner on every corner you're looking at about 30k polys on the low poly.


Yes it's pats acres go karting track

 

Yeah well known track in the drifting scene

 

Well my kart track is 6k triangles. I don't do an overall subdivide/smooth. I build on the spline and I subdivide on sections such as turns.This is only half of the track. The main track is only 2k triangles. The kart track has tighter turns so needs more detail to look smooth.



Again I work in blender so not sure how it works in 3DS. But I select the edges where I want it to be smoother and subdivide just that section. When it is still on the spline it will simply follow the spline better when you give it more resolution. This shot shot shows it right before and after I subdivide. Notice how just the edges are selected where I needed more detail.

 

I'll just add a one iteration mesh smooth on the visual then done 5 iterations on the physical. That should do it it's going to be really awkward to add a load of edges in and get them all to line up I've tried it's all over the place.

Here's a picture with 1 iteration of mesh smooth.


Then for the physical Ill do 5 which pushes me to 350k-400k polys in total.

 

The turns are still not smooth enough and you don't need to subdivide the width usually more than once.

Your turns will look like this


Where it should look like this

Seems nobody wants help or advice anymore...

 

1 metre spacing at the centre line will nearly always cover polygonal shape issues ive found. play with splines too because there is a method that will actually tighten your spacing where needed and stretch it out where you don't.

 

I'm just confused at the minute, blender must totally be different to max you're confusing me when you're saying about sub dividing and all this spacing etc what do you mean? I always thought subdividing in max was mesh smooth and turbo smooth.

I can add edges and swift loops but as i said lining them all up is a pain and it'll probably come off worse. And if i add mesh smooth just in the corners as i said I'm in high poly territory for a low poly.


And i am using a spline, i made my track line went across the segments separately and used the sweep tool as i went along.

 

Here's the spline.




Method I'm using.




Swift loop/connect edges

 

This tutorial makes more sense to me. Been a long time since I've used 3DS max and let me tell you I'm glad I learned blender for track making. Around the 1:50 mark he adjusts the path steps. I played around in max a bit to see if I could localize a subdivision like I showed above in blender but I couldn't find it. Worst case you do the whole road as it won't kill you.

 

you have way to many Bezier points there Ness. its possible with just 3 for a corner and none for straights. that way you can get the apex right over 1 handle. 3 for a curve and 1 on a straight (if required) 1 for entry 1 for exit and 1 for the apex. if its a dual apex then use 2. and that's the rule and that's exactly how tracks are designed from cad.

 

Ahh I can edit those points and re do the corners easily enough.

Now trying to find the best way to make the curbs at the moment, it's really hard to line up with the track and if you drag the edges out it's hard to get the shape.

 

That's lofting I wasn't keen on that sweep gives you more control however when I did use sweep I deleted the edges to make it lower poly so maybe I should keep em in.

 

Is this slightly better picture below, anyway i was going to just throw iterations of smoothing on it and leave it as one layer, but as you said you need a physical and a visual. So i'm guessing the visual is obviously what you are going to see in game as the physical is invisible so as you said the corners need to look like corners.

What i did this time is use sweep and kept the edges in.

That's the only turn I've done so far the first one I'll do the rest tommorow.

 

Better but still needs more subdivisions to look round. If you can tell it is a bunch of straight lines it isn't dense enough.

I know you understand the visual physical thing but just realize the numbers we are dealing with. Say your visual road has 10k polys. Your physical road will be 400k polys. So Don't skimp on the visual especially on turns. The tighter the turn the more you need. Hell my kart track has more polys than the main track and it is just for show. But being it has such tight turns I needed to add the detail to make the turns smooth.

 

I can only speak for my self, but I am holding on to the Goldnuggets from You. I have bought Asset Sketcher from your recommendation and have begun dabbling in the use of splines.

Texture baking is temporarily on hold, because of information overload.

On top of that is the RBR Rally lessons. They have been following me like nagging ghosts, ever since I learned that it is possible to create Your own tracks.

The fruits of Your help and advice will show in the future.

It is the very much appreciated effort from You and others, that helps Noobs like me, to create better results.

I know that I am going to need more help and advice from You.

I hope that I in time, will be able to help You!

Happy modding

 

All the Max tools are a pain but you get there with random combinations based on what you're doing.

Oh for a fully non-linear approach with clever systems for looking at edge angles for smoothing bends, but with min/max intervals etc.


But there is no doubt that at some point (with max's tools at least) that you have to collapse the stack and there is no going back without losing work.

This is where heavy prototyping and testing are essential so you never need to go back, at least not so much you're reworking everything!

As a point of reference, I did 20km if road with visual mesh at ~50cm grid size draped over laser scan, with a physics grid at ~ 1cm.

I'll be honest the 50cm grid felt good to me.

Why AC didn't add a nice perlin noise system with a texture based amplifier is beyond me.
For most non-laser tracks those high-def generic noise meshes are just a waste of disk space imo.

 

You can do just that in blender with the displacement modifier. When done properly, simulated noise can feel just as natural as the real thing.

 

The issue was more that you're essentially using data that could be in a LUT (texture), or procedurally generated at contact patch * world coords, encoded in a repeated manner across an entire track in memory.

It just seems a bit weird.

Of course with lidar and real fine topography then sure, but just faking it (despite it feeling good) via a method optimised for lidar is just a bit wasteful imo.


All being said, I like the idea that with this control you can in theory paint various real noise patterns (caught with photogrammetry, laser etc) onto your dense mesh and get a nice result with variability, lumps, painted line feelings etc.

But chances are most people add simple repeating perlin noise in Blender or Max etc, so why not just add that into the physics engine tyre z values and save millions of polygons and author effort adding them?!

 

I'm not a game designer but I do know nobody is doing such a thing. Maybe you have a ground breaking idea or maybe they have all tried such a thing and it failed, who knows. But what I do know is the track has to be fixed for things like 3D grass to line up with the track / terrain. The visual mesh must match the physical as close as possible so the car doesn't go in and out of the track.

Also it isn't very hard to make a physical mesh. Plus it isn't rendered so it doesn't effect performance in any way. And finally a 48MB file for physics really isn't that big a deal this day and age.

Oh and as far as I know the concept behind AC was supposed to be all laser scanned tracks. Obviously that has changed over the years but I am sure that idea of all LS tracks had a bunch to do with how the game engine was designed.

 

Free Racer had perlin noise for a decade.
The deviations against the visual mesh would be no different if the noise is provided via a clone of the visual mesh + mesh noise, or just the visual mesh + the end result on the noise on tyre z position.


The only point AC doesn't have it is because they almost exclusively used dense real lidar physics mesh.

But for modders without that data, it is wasteful to just add generic noise.


But it offers plus sides for modders who want to go the extra mile.

Sadly for many it just adds a fairly complex step to add a nice grainy effect to the driving surface.


I'm neither pro or against it.

It's just a shame there is no built in noise for those who can't be bothered doing all that extra work which could be automated.

But I wouldn't want to not see it there either because I'm sure it could be useful for those wanting to convey various complex surfaces.