๐งต Untitled Thread
Anonymous at Tue, 18 Apr 2023 20:03:09 UTC No. 943440
Hello /3/. I'm modelling a concrete block and was wondering if there is an way to optimize this to a game fitting level, since this skyrockets my polycount to a high level. How would bake it or reduce the edges on it?
Anonymous at Tue, 18 Apr 2023 23:05:42 UTC No. 943465
>>943440
as the replies have said, games would traditionally use normal mapping and/or parallax occlusion mapping for this. If it HAS to be geometry, do it like so. green are areas to be solidified, red are new edges to be added.
Anonymous at Wed, 19 Apr 2023 03:20:59 UTC No. 943485
>>943465
It has to be full geometry, can't pull off the 2D texture on.
>>943468
>>943469
This one is good. 500/300 tris is much more better than the 1k I had, I'll try this out. Thanks
Anonymous at Wed, 19 Apr 2023 03:24:29 UTC No. 943486
>>943485
To not give it a blocky appearence, since I'm attempting to go for a realistic level, should I chamfer the edges on the patterns aswell as the outside edges then use it as a HP mesh to bake the textures on?
Anonymous at Wed, 19 Apr 2023 04:39:31 UTC No. 943489
>>943486
If you're going for realism, you should subdivide it, sculpt imperfections, then bake normals/bump/cavity/etc to your low poly.
Anonymous at Sat, 6 May 2023 06:41:06 UTC No. 945201
>>943486
you should make a trim sheet
Anonymous at Sat, 6 May 2023 14:27:44 UTC No. 945218
>>943469
You could go one step further and remove the 16 polys on the sides that would be hidden when lowered in the bigger object, thus getting to 282 tris.
Anonymous at Sat, 6 May 2023 14:31:43 UTC No. 945219
>>943440
Is that an anti-homeless tile?
Anonymous at Sat, 6 May 2023 14:58:22 UTC No. 945222
>>945218
That's supposing he (you?) had those faces set to begin with. In such a case, you could possibly get four or 5 more depending on how the base was constructed, though that could easily be none if it was done efficiently to begin with.
Anonymous at Sun, 7 May 2023 02:12:35 UTC No. 945277
>>943468
Am I tripping, or is your edge arrangement more costly? With the "floor" set up in squares like OP has it, I count 66 tris. With the floor set up as trapezoids the way you have it, I count 96 tris.
Remaking OP's thing with square floors, I get a total of 482 tris.
Anonymous at Sun, 7 May 2023 04:57:39 UTC No. 945286
>>945277
You're right, actually. I dunno if that poster suggested that arrangement for a different reason or if it was a convenience choice.
Anonymous at Sun, 7 May 2023 08:37:08 UTC No. 945292
14 replies for a question that is infinitely stupid in the first place, with one answer that doesn't understand the assignment and another that is as incompetent as OP before the 14th reply finally has someone who grew a brain. I fucking love /3/.
Anonymous at Sun, 7 May 2023 09:13:14 UTC No. 945295
>>945292
>Counts replies
Hey, thanks for saying I have a brain. Incidentally, I forgot to clean the corners. I could have saved 8 more tris. There might be some reason for the trapezoidal method. Maybe it shades better or something. I don't know. I'm inexperienced with 3D.
The floating geometry method is still the most optimal way I can see. Cutting out the unseen faces, it comes to 282 tris.
Anonymous at Sun, 7 May 2023 12:42:38 UTC No. 945304
>>945295
Actually, now you've 'dirtied things up', so to speak, as you've introduced ngons. Doesn't matter too much for a rigid body and it'll all get converted to tris afterward, but your savings will disappear accordingly, just so you know.
>>945292
You're the worst reply in this thread by a long shot. Might be this kind of useless post that just shits on people without substance that makes the board so slow.
Anonymous at Sun, 7 May 2023 19:06:49 UTC No. 945337
>>945304
No Ngons here. All quad faces.
>>945308
I don't know what T-junctions are. But I suppose it would be a result of the geometry actually connecting at the points you highlighted. But they're not connected. It's an illusion. They're separate parts, flush up against each other.
Anonymous at Sun, 7 May 2023 19:29:18 UTC No. 945338
>>945308
I should also mention, that it wasn't under Eevee before. It was solid view with wireframe enabled. Here, I turned off the wireframe and turned on Eevee. Gave it a gray material. This is what it would look like. Well under a single white point light anyway.
I don't know how a game engine would handle this though.
Anonymous at Sun, 7 May 2023 19:40:18 UTC No. 945339
>>945337
T-junctions are vertices resting on edges. Due to limited numerical precision, you can't expect those to close perfectly. They're sometimes referred to as "word sparkles" because when there are T-junctions and you have a bright background, you can sometimes see through the object.
Anonymous at Sun, 7 May 2023 19:41:41 UTC No. 945340
>>945338
>It was solid view with wireframe enabled
It's still Eevee. Workbench is also Eevee.
Anonymous at Sun, 7 May 2023 19:46:10 UTC No. 945341
>>945340
oh ok. I'll keep that in mind. But the wireframe causing weird anti-aliasing. That was probably just because the parts were separate. So it was two wireframes overlapping.
Anonymous at Sun, 7 May 2023 21:08:20 UTC No. 945344
>>945277
Anything under 1000 triangles is a-ok for mobile. Why do you care so much? Better worry about texture optimization
Anonymous at Sun, 7 May 2023 21:12:07 UTC No. 945345
>>945344
I was just randomly browsing /3/, and saw a thread that even an inexperienced person like myself can contribute to. Learning how to optimize topology is pretty fun. It's like solving a sudoku.
Anonymous at Mon, 8 May 2023 01:20:31 UTC No. 945356
>>945337
Fair, then. Have you cleaned out any unnecessary quads from beneath the base?
Anonymous at Mon, 8 May 2023 02:34:26 UTC No. 945359
>>945356
Of course. It's just one flat square. There wasn't anything down there anyway. I'm not the OP. Looks like OP is building this on top of layers. But nah, no layers here. Just flatness below
Anonymous at Mon, 8 May 2023 09:00:39 UTC No. 945400
>>945339
I've agonized a bunch about the imprecision of the sphere primitive when looking at the dimensions and finding some kind of '.000000001' displacement breaking symmetry. I don't know if that's some kind of floating point issue or whatever, but it was a nuisance, as I could sometimes get it to disappear while it would otherwise return after some modeling. Comes back with mirroring as well in some cases despite the other side or quadrants not having the issue.
Anonymous at Mon, 8 May 2023 10:12:28 UTC No. 945401
>>945400
Almost all calculations in Blender are stored as the single precision C float type, which is 32 bits wide. That was the common choice in the early 2000 when Blender started.
Very small errors are due to the fact that IEEE float types are based on a power of 2 system instead of a power of 10 system which is used by humans.
Anonymous at Mon, 8 May 2023 11:03:32 UTC No. 945403
>>945401
Thanks. I was just making a fairly uninformed assumption by association based on my limited knowledge of types. I appreciate the explicit explanation.
Anonymous at Fri, 19 May 2023 21:52:23 UTC No. 946247
>>943440
Just a flate plane, do it with texturing, use normal/height/bump maps