Page 1 of 1

samples vs super vs filter

Posted: Mon Feb 23, 2015 6:57 am
by KillPixel
After several experiments and comparisons I'm leaning heavily toward using filter. However, the wiki says not to use it because it "doesn't play nice with a lot of the more interesting effects".

I haven't noticed anything particularly weird (bleeding, maybe), what issues does it cause exactly?

Which method do you guys prefer and why?

Here are my light switches, just in case you see one of those "more interesting effects" that don't play nice with filter:

Code: Select all

-threads 8 -v -cpma -bouncegrid -bounce 100 -bouncescale 1.5 -dark -dirty -dirtscale 2 -filter -gamma 1.1
Minor tangent: I like the look of samples in some cases, and filter in others. It would be nice to be able to flag specific lights to be either sampled or filtered.

Re: samples vs super vs filter

Posted: Mon Feb 23, 2015 7:42 am
by Eraser
-bounce 100 ?

Doesn't that make it take ages to compile a map?

Re: samples vs super vs filter

Posted: Mon Feb 23, 2015 8:18 am
by AEon
I use:

Code: Select all

-light -v -fast -patchshadows -dark -samples 3 -bounce 8
Though it just occurred to me I may have forgotten to turn off -fast... hmm...


-threads 8

Used to use that too... but it turns out that q3map2 automatocally finds out how many cores and hyperthreads are present under Windows 7 (64bit), making the switch redundant.


-bounce 100

Hmm... did you ever check how many bounces the light stage does? In my detailed maps it usually only does 4 bounces and has nothing to do at the 5th one... so 8 bounces is plenty.


-dirty

I had some issues with this in AEdm7, when combined with the shaders used from Sock's Plants (from Pyramid of the Magician)... so if you use models with more complicated shaders on them, take a closer look. And I seem to recall vaguely that -filter, also caused some issues in this context... alas don't recall what exactly.


-gamma 1.1

I don't recall messing with gamma ever actually improved the quality of the lighting. It made everything mushy and undefined. But some of the artists here, Hipshot IIRC, actually were able to brighten their maps *and* make it look good.


-cpma -bouncegrid -bouncescale 1.5 -dirtscale 2

Never used any of these.


In Hipshot's and my SolarAE we used:

-light -v -fast -patchshadows -samples 3 -bounce 8 -gamma 2.5 -compensate 3

Note the compensate to help with the usage of gamma.

Re: samples vs super vs filter

Posted: Mon Feb 23, 2015 8:19 am
by Hipshot
Bounce stops when there's no more information to bounce. I use 100 too.

Re: samples vs super vs filter

Posted: Mon Feb 23, 2015 5:39 pm
by obsidian
The Q3Map2 wiki is largely written by some end user of Q3Map2. It's an okay resource, but I wouldn't take it as fact.

-threads isn't used on Windows. As mentioned, Q3Map2 detects the number of available threads. It's only useful on Linux (and OS X?).

-filter is just a Gaussian blur effect applied to the lightmap. There's nothing wrong with this, but it can over-blur some shadows and cause them to lose detail. -samples on the other hand, was specifically designed to remove aliasing on shadow edges, so it's a smarter algorithm and should result in better results.

Re: samples vs super vs filter

Posted: Tue Feb 24, 2015 1:49 am
by KillPixel
Hmm... did you ever check how many bounces the light stage does?
It varies, I just use 100 to make sure I get all of them.

In my test map, the increased gamma and bounce scale with the increased AO has a pretty cool effect. The gamma increase is only .1, most of the extra light is provided by the bounce scale.

Haven't done a comprehensive test with -compensate (I have no idea what this actually does), that's next on the list.
The Q3Map2 wiki is largely written by some end user of Q3Map2. It's an okay resource, but I wouldn't take it as fact.
thanks, good to know.

Maybe if I feel really freaky I'll just use external lightmaps and process them in photoshop...