Oops, that should be fixed now with commit 9d3f1c7c9.

Ok getting there :)
Now it seems to try to tile the specmap wrong, as well as there being splits where there shouldn't be, and the edges of maps seem to become fuzzy.

Also, it doesn't obey 0 specularity on alpha channel 0 (unless I'm mistaken in my interpretation and this is desired behavior.

The stripe on blackalphachannel.jpg is a bug I'm working on now; what are the dimensions of your spectex image? [edit: resolved]



That was a bit of intentional laziness on my part, gone in git.



You have to set the RGB values to 0 if you want zero specular contribution, not the alpha channel.

Currently its 2k*2k for an 8k*8k texture map. Seems to be high enough rez.

Ah, so alpha only defines how "polished" the surface looks, and and the color is the amount of light that gets reflected for each color channel ?(added on the the pixels color values?)

Ill test latest git again, thanks!

neddies map would look great with that, if he evar released it..

lmao i love u man

Appropriate use looks great.

Which of the thus far posted screens do you count as appropriate?

Question (since I've been a bit busy with some other things lately): how does this differ from pre-lighting the map? Do the specular highlights change as the camera moves? Still thinking that tiled normalmaps for detailmaps would produce a really superior effect, myself, if the specular factor was controllable (to prevent the "crinkly plastic" effect) but I am curious as to how the big-rez maps are working- I'd have thought at one-quarter resolution to diffuse rez, it'd be too blocky to really work well, but the screenshots certainly look promising

Your iceberg map. Although I'd have to see it live to be sure that it looks as good as it appears to, static images have the ability to create false contours.

Kloot, I LOVE you!
Thanks for plugging in the shader, Im going to go nuts with it, since now its finally time to learn GLSL!

It differs because you can't precompute highlights for every viewing position + angle. Lightmaps also don't play nice with terrain deformation.



Yes, that's kind of the definition of specular lighting.



all in good time



The ubiquitous plastic look mostly comes from the lighting model itself. There exist better, physically-based models, but none that are as cheap to evaluate as good old (Blinn-)Phong.




Fixed locally; I'll commit the patch tomorrow.

So, after a few more rounds of testing:
The tiliing screws up in different ways depending on the size of the speculartex. The spectex says
This
is an
SSMF
test

512


4096


2048


Also, the features get all red.

And there are these odd brown spots on the east slopes of hills (where there are shadows and are pre shaded to be quite dark)

I've pushed the spectex tiling patch upstream; you'll need to update engine and base content.

About those red features, do they render normally if you turn shadows off? If so, can you check your infolog for any shader errors?

Damn, my main systems hdd gave up, it will be a while till I get it up and running again. No data loss though :)

@Kloot: Thanks for answering my questions, this looks like it's going to be pretty damn awesome, will check it out as soon as I'm not buried under some other things.

Woot, now it works with spectexes under 1k in size, for a 2k sized spectex it does this:

THIS IS
AN
SSMF



Also, on a related note, i got the epicness to work:

No luck reproducing that one so far I'm afraid, see http://img94.imageshack.us/img94/8915/ssbv2ssmf.png. There is no tiling with any of the attached three textures (2048x2048, 4096x4096, and 8192x8192 RGBA PNG's respectively). May be a driver bug or texture size limitation (although modern GPU's should handle even 8K textures just fine), but I'm not sure.

Dunno, drivers are garden fresh, gpu isnt all that old (8800gt), ill try more to reproduce it. Nothing in the shader logs either.

What do you mean, "epicness"?