 News | Forum | People | FAQ | Links | Search | Register |
Log in |
|
|
Which topic has the most unread posts on your profile. Mine is "other games" which is 8072 new. Silly func, implying there are games other than quake.
Any Quaker in Berlin who'd be interested in meeting? I'm here for vacation until Friday and I don't have a lot on my schedule. Email address in profile is up to date. Beer is always an option!
I had my computer take a closer look at the contents of palette.lmp, since I wondered whether all the color values are unique. It turns out they are not and there are twelve duplicates.
I show my work here for those who care how this information was obtained: http://pastebin.com/tSaGppSX My result is that the following twelve pairs of colors in the palette are the same. The indices are given first, and start from 0. 0 0x000000 48 0x000000 112 0x0b0700 207 0x0b0700 32 0x0b0b0f 223 0x0b0b0f 143 0x0f0707 159 0x0f0707 16 0x0f0b07 175 0x0f0b07 51 0x131300 80 0x131300 142 0x170b0b 158 0x170b0b 52 0x1b1b00 81 0x1b1b00 141 0x231313 157 0x231313 53 0x232300 82 0x232300 76 0x670000 247 0x670000 111 0xfff31b 192 0xfff31b Indices larger than or equal to 224 are normally fullbright, so this means one color, 0x670000, appears as both non-fullbright and fullbright color and thus the two instances are not technically the same color. Other pairs are made up of two non-fullbright colors which just happen to appear in different places in the palette. Having 0x670000 be a color of both kinds appears to make automatic conversion of full color images to Quake textures problematic. How can a program tell if a color in the original image that maps most closely to this color is meant to be fullbright or not? You have to pass the information about which pixels should become fullbrights by some other means, I guess. This may be old news to some folks here, but it was new information to me. I don't know why there are duplicate colors in the palette or if the pairs are handled differently by any engine. If anyone can add anything to this, I'd be interested. There is no benefit to any of the dupe colours. Most likely just how whoever made them first in DeluxePaint set up some ramps to go from the same colours to other colours, or copied some colours around. The Doom palette also has 7 dupes I think.
And yeah, that fullbright red colour is seen a lot in some converted texture.wads, esp those compiled in the hayday of GLQuake where you would not see fullbrights at all. The only solution is to ignore fullbrights when importing, or to go over the image in case there are parts you want to be fullbright, to ensure that those parts are and other parts are not. Another really funny thing about the Quake palette is how the first 8 rows of 16 colours go from dark to light, but the next 6 go from light to dark. These 14 rows are all used as swappable colours for the shirt and pants graphics. And because again, one of the artists who made the palette fucked this up, for no appearant reason, John Carmack hacked in a fix so that the palettes which have a different brightness order would not look inverted on the player model. Fudge abound :D There is no benefit to any of the dupe colours
Actually they're quite important in one particular of the quake engine, and further down the post you even mention it! They make the shirts/pants system work, the colours that are duplicated appear on different rows, and you'd need to replace the row based gradient system with something much more complicated to eliminate the duplicates and still have the same visual on the players. I also checked the colormap.lmp for duplicate color columns. If my quick Python script is not wrong, the eleven identical non-fullbright color pairs have identical columns between them, but obviously colors 76 and 247 cannot have the same columns because the latter is fullbright. There are no other identical column groups in the file.
As I understand it, only software-rendered Quake engines use the colormap. Not really. It would not have been hard to slightly vary those dupes up and as a result get a slightly smoother colormap plus no ambivalence when repalettising.
Here is a tweak I just did which should not have dupes (unless if fucked up slightly, which would not matter, the point stays the same.) https://dl.dropboxusercontent.com/u/15588722/post/func_msg/palettetweak.png I converted the palette to 18 bit colour before doing the edits, since that is what VGA mode 13h could output afaik. (albeit stretched to full 24 bit space in this case) Not really. It would not have been hard to slightly vary those dupes up and as a result get a slightly smoother colormap plus no ambivalence when repalettising.
But you end up either changing the colours in an un-noticable way, in which case why bother - or you change them in a noticeable way and people would notice. I'm not convinced that making the palette a bijection has ever been on the mind of any designer. Here is a tweak I just did which should not have dupes (unless if fucked up slightly, which would not matter, the point stays the same.) I rest my case...
|
|
| You must be logged in to post in this thread. |
|
| Website copyright © 2002-2024 John Fitzgibbons. All posts are copyright their respective authors. |