Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Thu Feb 01, 2007 12:28 am Post subject:
Voxel Section Editor III 1.31 released!
Subject description: The normalizer won't be updated anymore...
Hey guys, if you thought the curves of the voxels in the news below were perfect, you have yet to see what is 'perfect normals'. Download the new Voxel Section Editor III 1.31 and take your conclusions. The normalizer is the only change of this version.
The same Mad Dog with the new normals. A bit darker, but clear curves. QUICK_EDIT
Joined: 25 Aug 2004 Location: DAS BOOT IM DER OSTSEE
Posted: Thu Feb 01, 2007 11:49 am Post subject:
I downloaded it and I have the same normals options as older versions. Smooth Normals and Autonormals. _________________ PPM's Reichstrollfuherer, 236th Trollenparties brigade. QUICK_EDIT
Joined: 22 Aug 2006 Location: somewhere south of the north pole
Posted: Thu Feb 01, 2007 12:45 pm Post subject:
Looking good, looking good, testing later today. BTW, can I link to teh file threw my own site? I felt I needed to check with you _________________ This is a signature QUICK_EDIT
I thought it could be nothing but perfect, until I tried it on my recently made (although still WIP) MLRS, and I saw weird black stripes going around the voxel. I have no idea what would cause this error.
MLRSbug.PNG
Description:
What is even more wierd, is that the stripes are visible only if you look at the voxel when it's pointing SW - SE.
You either have a hole in the voxel, or you have some 1 voxel wide details that can't possibly normalize properly. The former sounds about right. _________________
Joined: 22 Aug 2006 Location: somewhere south of the north pole
Posted: Fri Feb 02, 2007 9:32 am Post subject:
Desu wrote:
I downloaded it and I have the same normals options as older versions. Smooth Normals and Autonormals.
Yes, that's correct, but didn't you notice that when you used autonormals, it did say Autonormals 4.1?
I did some tests, and i have found it doesn't like sharp edges very well... But if you smooth the edges a tiny bit, it looks lovely _________________ This is a signature QUICK_EDIT
I filled the voxel, but those stripes keep appearing. It has no one pixel-wide objects.
Here, I made an animation, I hope it clears up things.
Now that I made the animation, I noticed that the stripes are not as dark as in the VSE III...hmm...
EDIT: I see what the problem is: the normals doesn't normalized sharp edges correctly, like Dupl3xx said. I tested it on many voxels, and I even made a box, a totaly sharp - edged box, and you know what? the normals were as shit as...shit.
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Fri Feb 02, 2007 8:50 pm Post subject:
I've seen this problem in another voxel and I've made a way that could fix it (by setting all normals to positive). However, it screwed up almost every other voxel making one of the sides totally dark. QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Sat Feb 03, 2007 4:50 am Post subject:
Ok, guys. While my tecnic is good at tracking the best normal value, I've noticed that there is a serious problem that may make it give the wrong signal to the normal (resulting in the evil blacks). The problem is very complex and happens when the difference between the normals behind it and in front of it (in any direction) is zero and I'm trying to find a way to deal with it properly. QUICK_EDIT
You can get 1.2c [url=http://www.d2kstudios.com/downloads/Command and Conquer|Category/Utilities|Subcategory/VXLSE_III_1.2c.exe">here[/url]. _________________
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Sun Feb 04, 2007 7:06 pm Post subject:
At the moment, yes. I already have a plan to make a way to get over it. However, how to implement this plan is still a mystery.
The Cubed Normalizer actually does a flood and fill in the voxel to find out what is inside and outside the voxel, but these 1 pixel wide sections might still cause the algorithm to find normals like (0,0,0). So, I have to solve the problem of what happens when the limit of the coordinates x, y and z tends to 0.
To fix this, I need to make it find a convex shape of the whole voxel and fill it with a nearly null value (like 0,0000001) before doing the normalization process. Last edited by Banshee on Wed Jul 25, 2018 10:12 pm; edited 1 time in total QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Fri Feb 09, 2007 8:09 pm Post subject:
Update:
I've been working in the concept of 'influence area' to sort the problem mentioned above. Today, I've tested the new implementation in a voxel that used to have this problem and it worked very well.
So, 1.32 should be released in less than 24 hours.
I'd like to point out that the normalizer does an almost perfect job as long as the user respects the mathematical limits imposed to the concept of voxel (the light reflects to only one direction, on each cube). So, there is no perfect normalizer (even if you export your voxel with 3ds2vxl) that will make your voxel perfectly normalized with one pixel wide shapes. Make it at least 2 pixel wide, to assure a good performance. QUICK_EDIT
The voxels I showed in the pictures had no one pixel wide objects, no antennas, no rails. Both of them had, however, sharp edges. I tested it and those black rails don't appear if the edges are rounded...well, it does, but a lot less. _________________
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Sun Feb 11, 2007 2:50 am Post subject:
There is a little delay in the version 1.32 because of the shortcuts that people asked me. Applying some of them is being harder than what I originally imagined. QUICK_EDIT
The normal of each cube is the lighting vector, DeathRay.
I know but... Say, you had a cube, if you average the possible directions, you would get a 45' angle at the edges, relative to the sides, and a 90' angle on all of the voxels on the sides. That's what I mean. _________________
Sorry for bother you by re-opening this thread, but I had urgent thing to discuss. Unlike you said Banshee, after downoading the old 1.2c and normalizing the voxel I presented above, the results were great, just like they were before. No stripes, nothing. _________________
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Fri Feb 16, 2007 4:54 pm Post subject:
In the demo truck, V3 looks better than the newer ones. In the orca, 4.1 is apparently the best. But 5.0 is better than 3.0 in this case.
As far as I can see, I can make the value that tends to 0 smaller to make sure it reduces its influence in the normals calculation of the sectors that are at the border of the shape... QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Fri Feb 16, 2007 8:07 pm Post subject:
Actually, re-looking the demo truck... v5.0 looks better than 3.1. The back is more rounded... and the difference in the front is more a matter of angle..
I've also tested both voxels with the new autonormals v5.1 and both are looking good. QUICK_EDIT
I forgot to mention that the rail at the back of the demo truck goes black when using 5.0 normals, and the remapped parts of the Orca goes black in 4.1 and 5.0. I still prefer using 3.0, because so far, it gives good results on all voxels. _________________
You cannot post new topics in this forum You can reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum You cannot attach files in this forum You can download files in this forum