Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Wed Sep 05, 2007 10:47 pm Post subject:
Range 6 is kinda high for Normals 8.0, making it too much smooth. For normals 8.0, I suggest you to use range 2, 3 or RA2 (3.69). Less than that is horrible. More than 5 is usually twisted. QUICK_EDIT
Joined: 22 Aug 2006 Location: somewhere south of the north pole
Posted: Thu Sep 06, 2007 7:40 am Post subject:
I think AN 5.6 with 5range and smooth on, is BEST! Ant to banshee, AN 8 is BAD! tested it on a 10x10x10 cube, it hurt my eyes And AN 5.6 with 5range bets the h3ll out of AN 8 with RA2 range... I don't know what can be done, sorry... _________________ This is a signature QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Tue Sep 11, 2007 4:20 am Post subject:
AN 5.6 is a way too much smoothed. AN 7.1 fools your eyes. The lighting angles are wrong, but the result is consistent. AN 8.0 is better used with range 2. The results are more realistic, but there are some chances of the oposite normal be choosen, which makes you see a black dot instead of the correct normal. I'll try to work in a way to reduce this problem... QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Wed Sep 12, 2007 3:57 am Post subject:
Hey guys. I managed to track some bugs in the ray casting part of AN 8.0 and the results were heavly improved in my end. I should commit the changes soon to the SVN.
Joined: 22 Aug 2006 Location: somewhere south of the north pole
Posted: Wed Sep 12, 2007 5:16 pm Post subject:
hehe, now that would not be the most impressive done by vista yet I now need the vista disc to be able to start Am I just going to replace the file or? Did try replace, but it gave me an "This file is not a valid Win-32 program"
Am I doeing something waay wrong, or is just vista being a bad witch?
EDIT: Also, the new file is ~100KB and VSE III .37 is ~1700KB Don't tel me you only complied the new AN 8.1, not the rest _________________ This is a signature QUICK_EDIT
Joined: 22 Aug 2006 Location: somewhere south of the north pole
Posted: Wed Sep 12, 2007 9:23 pm Post subject:
I see, opera didn't like it :'( But there is always IE7 at the back of the bus
EDIT: It's better, but I'll still use 7.1 and 5.6
AN 8.1 is very good at some points, but overall, meh
I don't know, I havent got the time to test every single on ingame, but VSE gives some "good" signs...
_________________ This is a signature QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Tue Sep 18, 2007 1:18 am Post subject:
Smooth normals doesn't mean that it is good. In fact, it is bad, because it doesn't detect the borders correctly. Good normals are smooth when it needs to be smooth and detect the borders in a proper way, which AN 5.6 doesn't do. I'm improving the border detection ability, while the smoothness can still be reached with a higher range. QUICK_EDIT
Joined: 22 Aug 2006 Location: somewhere south of the north pole
Posted: Wed Oct 10, 2007 9:42 pm Post subject:
Sorry the bump, but this is important! If you run ctrl+q you will get very good normals in an 8.1 but many of you know hollow voxels looks bad ingame _________________ This is a signature QUICK_EDIT
First of all, it is much faster now, since it no longer normalizes voxels that are inside the volume. The second thing is that it no longer normalizes 1 pixel walls or lone pixels, which are mathematically impossible to normalize. So, you'll notice a fake feeling that the normals are better or not at these regions. There is also some little improvements when dealing with other special cases, which really improves the final quality of the normals. Check it out and have fun.
During my tests, they've got excellent results for everything, from Kirov, passing through some awesome cruisers to Duplexx's challenging voxels. QUICK_EDIT
The old normals kept producing on my voxels either 1-black spots, or 2-weired-shaped patchs on the flat surfaces.
I really hope these auto normalizers are better! QUICK_EDIT
Normals=good
Bugs=Bad
It caused my PC to restart for no reason and I had to disable then re-enable the 3d window to get it to work each time... _________________ Yes, work on MV continues. It is not forgotten. QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Mon Dec 01, 2008 10:55 pm Post subject:
There is something I forgot to say: I'm using windows xp 64 bits to develop it. I'm not sure if it is compiling with 32 bits or 64 bits... :S.Considering what you said, I'm not surprised if this is a 64bits thing. I'll try to figure out how to compile in 32 bits here. QUICK_EDIT
There is something I forgot to say: I'm using windows xp 64 bits to develop it. I'm not sure if it is compiling with 32 bits or 64 bits... :S.Considering what you said, I'm not surprised if this is a 64bits thing. I'll try to figure out how to compile in 32 bits here.
Not possible. If you compiled it in 64 bit mode, he wouldn't be able to run it at all! Windows would not recognize it as a Win32 executable. _________________ QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Mon Dec 01, 2008 11:39 pm Post subject:
Yea, just checked Delphi help files and FAQ and my Delphi really compiles in 32 bits. What happened to Lt Albretch's computer is beyond my scope. QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Tue Dec 02, 2008 9:14 am Post subject:
You just used a model with a lot of thin (1 or 2 pixels wide) planes, which is exactly the weak point of working with tangent planes, as well as the treat discontinuity option that I'm trying to fix here.
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Wed Dec 03, 2008 2:35 am Post subject:
I've improved the thing once again. Click in the link of the post above to grab it.
Banshee wrote:
More improvements on auto normals. For those wondering why I'm working so much on it. I've noticed that tangent planes failed with 1, 2 pixels wide surfaces. You cannot generate a tangent plane with only one pixel. And a tangent plane with two pixels is hardly reliable. So, for these cases, the tangent plane mode is now using the Cubed Normalizer method, which works very well in these situations and awful on bigger scenes. I'm still trying to figure out a nice way to mix both of them to get the best scene on every case.
You can 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
Normals test
And AN 5.6 with 5range bets the h3ll out of AN 8 with RA2 range... I don't know what can be done, sorry...
I now need the vista disc to be able to start 
D'oh!
.
