[Danukem]

CraigFatman, on Jan 10 2011, 12:08 PM, said:

This is the truecolor version of palette #2 generated by an incomplete version of my script (that is, no manual editing). I've resaved it to PNG to reduce file size. It seems to fit HRP textures quite well. Other palettes do as well, but I'm not satisfied with ones which preserve most of the colors unchanged. I'll try to refine my color mixing algorithm to get something like a fuzzy Voronoi diagram.

Hey, it would be fantastic if you could generate something for your CLUT pack to work with the HRP.

[Jan Satcitananda]

DeeperThought, on Jan 10 2011, 11:16 PM, said:

Hey, it would be fantastic if you could generate something for your CLUT pack to work with the HRP.

Highpalookup maps are capable of far more complicated color conversions because they aren't bound to predefined palette entries. So, if I create a HRP version of ExtCLUT pack, it would comprise completely different filters. The majority of ExtCLUT palettes are just multiplicative filters which can be done in HRP by means of simple tinting (although I dunno whether it's possible to brighten a texture in this way).

[Danukem]

CraigFatman, on Jan 10 2011, 12:37 PM, said:

Highpalookup maps are capable of far more complicated color conversions because they aren't bound to predefined palette entries. So, if I create a HRP version of ExtCLUT pack, it would comprise completely different filters. The majority of ExtCLUT palettes are just multiplicative filters which can be done in HRP by means of simple tinting (although I dunno whether it's possible to brighten a texture in this way).

No, it isn't possible to brighten a texture that way, and it's one of the main problems with using tints. Also, while a majority of them are multiplicative filters, the exceptions are important. The point of having an HRP version would be that someone could play a mod that uses the CLUT pack either with or without the HRP and the colors would be similar.

[Plagman]

Here's a preview of Lezing's pal 2 applied in game:

http://picasaweb.google.com/pierreloup.gri...781364684118786





As I said in the last image comment, this is looking pretty much spot-on except we don't have a way of reproducing the gradual shift to red of pal 2 as it gets darker.

[Plagman]

Actually, we might. Let me try something.

[Plagman]

Yes, that's closer:



The shade scale just needs a little tweaking, but that's a more general problem with the mapping between shade offsets and OpenGL fog that I intend to tackle later in more detail.

[Danukem]

That looks really nice.

What would make my jaw drop would be seeing Duke with all his color variations (pals 10-23) without using any alternate skins.

[Jan Satcitananda]

What I found strange during my testing is that while sectors and walls use highpalookups, the models continue using tinting. Thus, I saw no difference on monsters and stuff...

http://img15.imagesh.../9753/pal23.png

Also check out the pal #23. It was generated using the same settings as the pal #2. One can see that only a small region of blue colors gets repainted into yellow (what corresponds to the palette colors), the rest is left as is, because it lies outside method's "region of confidence". It works well with most HRP textures, but doesn't affect ones which have more saturated blue (such as the access card sprite), or have significant variations in hue/saturation. Also there's a weird spot in green colors.

[Plagman]

Ah, yeah. Helixhorned also pointed that out the other day. I forgot to hook the system up to model skins, I'll try fixing that shortly.

[Hendricks266]

Do these work in Polymost? Is that even possible?

[Plagman]

Define "possible". They would work if Polymost was modified to be a lot like Polymer, yeah.

[Plagman]

OK, they should now work with model skins. As with regular textures, they override any tint but are overridden by alternate palette definitions. The model skin code is kind of clunky, so let me know in case I broke some subtle pal/def interaction with models.

[Plagman]

And I did the change to make the behaviour like the last shot I posted.

[Tea Monster]

Cool!

So what do model makers have to do to get this to work?

[Roma Loom]

So basically it's time to remove all the alternate pal textures since they have priority over highpal defs, right?

[Helixhorned]

r1757 breaks some keys like Alt-P for me in Mapster32. Go figure

Roma Loom, Tea Monster: patience

[Plagman]

Breaks it how? On walls, sprites? It seemed to work here, but I might have missed something.

TM: Right now we need to improve the generation scripts to end up with a set of maps. Lezing's approach yields very accurate results without having to finetune everything for each palette, but there are some problems. For example the nightvision pal is going to look way off if it only converts part of the space. We'd need an approach where the generation script is smart enough to extrapolate the regions of the base pal to cover all of RGB instead to get that right. Once we have a set of highpal maps, you just add them to the HRP and you're pretty much done, there's no work to be done for individual textures or models.

[Plagman]

Ah, I think editor aiming is just broken in Polymer because of my change, so that would explain what you were seeing.

[Plagman]

OK, should be fixed now.

[Hendricks266]

Plagman, on Jan 11 2011, 08:10 PM, said:

As with regular textures, they override any tint but are overridden by alternate palette definitions.

As far as I can recall, models have had tints override alt-pals, so if it works as you say, it is a welcome change.

[Danukem]

Hendricks266, on Jan 12 2011, 10:07 AM, said:

As far as I can recall, models have had tints override alt-pals, so if it works as you say, it is a welcome change.

Yeah, that would be a welcome change, and it should work that way in Polymost, too.

For example, in some of my mods, I have a purple tint for pal 17, because in many cases purple does kind of approximate the pal 17 sprite. But if there's a pal 17 skin, I sure as hell don't want a purple tint on top of it.

[Tea Monster]

Helixhorned, on Jan 12 2011, 03:44 AM, said:

Roma Loom, Tea Monster: patience

No problem, just keep us posted when you come up with something.

[Jan Satcitananda]

Plagman, on Jan 12 2011, 08:04 PM, said:

TM: Right now we need to improve the generation scripts to end up with a set of maps. Lezing's approach yields very accurate results without having to finetune everything for each palette, but there are some problems. For example the nightvision pal is going to look way off if it only converts part of the space. We'd need an approach where the generation script is smart enough to extrapolate the regions of the base pal to cover all of RGB instead to get that right. Once we have a set of highpal maps, you just add them to the HRP and you're pretty much done, there's no work to be done for individual textures or models.

Extrapolation is what I was thinking about too. I've made some progress with my script today, now I can release it. The algorithm is mostly empirical though, so it performs a large amount of operations and works quite slow.

http://lzg.duke4.net/clutstat.zip
This archive contains both the Evaldraw script and a set of palettes for it (##0-25 are standard Duke pals and the rest is ExtCLUT pals). You might want to test it with some HRP textures. If it succeeds, I'll write a page with detailed info for my site.

A few recommendations regarding color map generation:

The script has an option to process a limited range of palette colors. The last line (with 15 self-luminous colors and the transparent color) doesn't fit into the model, so it's disabled by default. The palette may contain ambiguous colors which can interfere with each other. For example, if you process a pal which changes only blue shades, you're advised to disable the line of violets as well (set the range to 0-223).

Default sampling level is 64, what stands for 6-bit precision. To attain 7-bit precision, raise it to 128 (it will take 8x more time to generate). To save a TGA image containing a highpalookup, hit spacebar to enter output mode and then save your image via F6.

The keyboard commands are:

Enter: rebuild the map;
Space: toggle output mode;
R: reload test image;
-, =: set upper bound of the color range;
[, ]: set lower bound of the color range;
Down, Up: set sampling level;
Left, Right: change active look-up palette ID.

[Plagman]

Any chance you can upload a set of generated maps too? It seems mostly straightforward, but what you said about disabling ambiguous ranges makes me think you would get the best results since you're most familiar with it. I'll give it a shot tonight, though.

[Jan Satcitananda]

Plagman, on Jan 14 2011, 01:28 AM, said:

Any chance you can upload a set of generated maps too? It seems mostly straightforward, but what you said about disabling ambiguous ranges makes me think you would get the best results since you're most familiar with it. I'll give it a shot tonight, though.

Ok, I've run the script for all palettes except #0, #3, #4 and #9; the result was temporarily uploaded onto my site (30 MiB): http://lzg.duke4.net/hipalset.rar
I've tested each pal in Mapster32 on various textures, seeing whether the palette set is consistent enough. It's not perfect, but gives satisfactory results for most tiles. Sometimes a tile contains colors which are very similar but lie in different palette sections (say, upper-red and upper-orange shades) and thus handled by look-up palettes in different ways what looks kinda weird. Beside that, the palette #23 distorts dark grays due to interference from blue colors being converted, and the palette #24 makes gray into pinkish red which looks less saturated than it should be.

[Plagman]

Alright, I gave them a try and the result looks pretty good. Here's the before/after:

[Danukem]

That is incredible! I want this right now. You might as well dump all the alt-pal textures and skins from the HRP, because this is sure to replace them when it's done.

[The Commander]

Yes, these do look very good and will save A LOT of space in the HRP and other mods that use alt textures.

[Danukem]

It is a huge deal. It saves a lot of disk space, and it will improve the look of the game when using the HRP. A lot of user maps and mods will be much better with the HRP now.

Big thanks to Helixhorned, Plagman and CraigFatman for doing this!!

[The Commander]

Now certain custom maps that used alt pals that had not been made yet in the HRP will actually look like they should.
(Thinks of maps by Gambini etc)