Because its Halloween or something...

Discuss the Wing Commander Series and find the latest information on the Wing Commander Universe privateer mod as well as the standalone mod Wasteland Incident project.
Post Reply
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

MK, I used to think the same way. I learned from experience. The first time I was working on the Demon texturing; actually, only two textures: diffuse and specular. I was just trying to come up with something quick for Privateer Ferrius to have as a starting point. Anyways, it was just two grey textures, 2K. I thought that because there were not too many other ships around, it would be okay.

It wasn't. At all. The frame rate was dismal. Choppy. I could hardly believe it, but there it was. So, I reduced the textures to 1K, and the FPS went up considerably. Then I decided to try 512 and it was smooth as silk.
So I don't care what anybody else says. I had my own experience.
And this was using a 6600GT gpu. Certainly two generations behind, but probably the fastest videocard anybody in the WCU team has.
Of course I had my FSAA and anisotropic filtering maxed out in the driver, and I had the resolution at 1280 x 1024; but I want those settings that way.

And really I don't see what's to be gained from going above 512 for a small ship. If you use 2k for such a small ship, what size texture should I use for the Bengal? It doesn't make any sense whatsoever. 512 should be plenty for a tiny fighter. The Caernaven's texture is 512, for Pete's sake.

If we're going to set standards, let's set standards that actually make a bit of sense, like say "an inch to a texel", and then "make sure we stick to it"... And we'd neve stick to it, anyways, for practical reasons: If a ship's surface was 1.5 million square inches, would you go to a 2k x 2k (4 million texels) and waste 2.5 million of them?

Every time I hear the words "standards" and "stick to it", they invariably come from someone who's not involved. By the way, you can do the texture hiding thing right now. In VS it's called subunits. Just divide the ship or thing into subunits and give a piece of the full texture to each subunit.

If we use the largest texture that the videocard can handle for the smallest ship, then what are we going to do for the big ships? I'll tell you what ends up happening, every frigging time: Eventually, all ships, big or small, end up having the same texture size, and you have no visual clues whatsoever to tell the relative sizes of things. Happened in PR, happened in VS, happened already in WCU. I'm trying to inject some minimum sense of proportion here. A small ship should have a small texture. Period. All you want to do is blow them up, anyways. For bigger ships, like Paradigms, we can go 1k. Let's leave 2k for stations, shall we?

There's no point putting on a bravado about "shooting for the stars" only to find yourself back on the dirt, heels over head. Before you even suggest things like standardizing to 2k, you should try it, first. I did. You didn't. So why are you telling me? Try it, and you'll see. It's not just about memory space; it's about performance too.

In any case, I spent like a week arguing with Brad at wcplanet, and now argument here again. I'm really tired of it sick. I'm out of here for a few days.
DualJoe
ISO Party Member
ISO Party Member
Posts: 387
Joined: Tue Mar 21, 2006 2:37 pm
Location: Netherlands
Contact:

Post by DualJoe »

Well since I'm walking around in my T-shirt up here in Holland and we're going south to the Alps I'm not too worried about freezing to death. Would be more of a challenge if we did it in your nick of the woods.

My health however is becoming a recurring annoyance. Currently my left eye has a scar on it and vision is reduced considerably compared to my right eye. What really worries me though is that once you have this virus in your eye it remains there and becomes a problem whenever your immune-system is weakened. Unfortunately a weakened immune-system is one of the many new features I got after my concussion. Pretty difficult to determine what changed after my little accident (pretty hard to measure yourself), but I do show signs of mild braindamage (loss of the sense of smell and taste, dizziness, problems with my metabolism, change in personality etcetera). I really have to think about how much I value my military-career over my own health.

Back on topic
Be sure to check out the path-tool in The Gimp. It's vector-based and will scale to any resolution. That's how I did almost all of the panels and rivets on the Hornet. Very easy to get the hang off and saves a lot of work in the end. You can trace over the lines with the brush of your choice to create lines and rivets. What I found the most useful apart from scalability, effectiveness and enormous control, is that it allows you to work on regions without needing to actually paint the lines. This opens up a whole range of effects like overlapping plating:

Code: Select all

__                         _____
   \            _____ -----      \
    \_____ -----                  \_____
Differences in height between certain surfaces to suggest opening panels (like the landing-gear-hatches on the Hornet) and stuff like the grills (the one on the back of the Hornet). When you copy this grayscale layer on top off the diff and let it merge it gives an extra sense of depth (and wear and tear) to your texture even without the fancy lighting.

Another option would be, because like me you seem to prefer modeling over texturing, to model details in Blender. Then with the use of the node-system render the z-buffer, radiosity/GI and diffuse and then paste it onto your respective gimp-textures. I've shown some examples in the last post in my hornet-thread. It's actually a lot of fun and allows for more interesting shapes than I'm able to paint in 2d.
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

DualJoe wrote:I really have to think about how much I value my military-career over my own health.
No kidding. You should do well in any career you choose. Modelling/texturing for a living is not out of the question; nor is getting into a game company venture. Changes of career get harder the older you get, so I wouldn't just wait.
Be sure to check out the path-tool in The Gimp. It's vector-based and will scale to any resolution. That's how I did almost all of the panels and rivets on the Hornet.
Rivets? I thought it was only for lines; gotta check this out.
effects like overlapping plating:
I had no idea you could use paths to select regions; sounds pretty useful.
I'll check it out; thanks.
Another option would be, because like me you seem to prefer modeling over texturing, to model details in Blender. Then with the use of the node-system render the z-buffer, radiosity/GI and diffuse and then paste it onto your respective gimp-textures...
I never even heard of z-buffer rendering; where is that? Sounds great.

Right now I'm struggling with normal bakes; I made one for the low poly mesh, one for the high, subtracted them and added 0.5,0.5,1.0 and normalized it with that Gimp script; and now I can't tell what gain to give it so I'm doing like a dozen renders to play them like a movie to see where the smoothest point is.

Gotta go to the bank; brb.
DualJoe
ISO Party Member
ISO Party Member
Posts: 387
Joined: Tue Mar 21, 2006 2:37 pm
Location: Netherlands
Contact:

Post by DualJoe »

OT:
You can't make much of a living in the modeling/texturing business over here in the Netherlands. Trying to get my bachelors at the academy is currently the best move. I'm working on some backup plans in case plan A fails (looks like it will).

Baking:
Seeing how many people are asking for the high->low baking feature I suspect it will be in Blender pretty soon. Till then I'd use the old BRaybaker script for most of it. For normalmap baking I'm using the dragengine normalbaker, mostly because I prefer working in Linux and secondly because it produces the best results with minimum effort.

I can't figure out how that multires-mesh feature works for LODs. Looks like you have to create the LOD first and then add resolution to get higher res. I doubt this will be very useful for making game-models. Maybe it can be used for baking highres details in the low version.

Rivets:
As I recall I made a soft-brush for the rivets and increased the spacing. When you're satisfied with the spacing you can use this brush to trace over the paths you selected.

z-buffer etc:
With the new node-system Blender can render multiple layers at the same time. This means you can render the z-buffer of the image to a gray-scale-image and presto you have a depth-map that can be copied onto your handcrafted bump-map.
Would be nice if we created a library of these details, so we could recycle them and get some uniformity across the models. In fact you suggested something along these lines in the Hornet-thread Chuck. I still have some samples I made there, awaiting wcu-dev crits.

I've just figured out another way to make this work. The new sculpt feature in Blender works with multi-res. It also supports custom-3d-brushes. If we use those pre-generated details as brushes you can literally stamp the details into your models. Then, because of multi-res, you should be able to bake the textures for the low-detail (the factual game-model) using a higher level of resolution.
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

DualJoe wrote: Baking:
Seeing how many people are asking for the high->low baking feature I suspect it will be in Blender pretty soon. Till then I'd use the old BRaybaker script for most of it. For normalmap baking I'm using the dragengine normalbaker, mostly because I prefer working in Linux and secondly because it produces the best results with minimum effort.
I sure hope Blender devs do that. In the meantime, xNormal is very much improved. The new version is only 3 days old and it shines. It gives you complete control now; if you don't want filtering just turn it off.
So, these are the settings I'm using:
Max frontal ray distance: 0.06 (Have to play with it, depends on the model)
Max rear ray distance: 0.02 (My coarse mesh is slightly enlarged along normals, though xNormal no longer requires that the fine mesh be inside the coarse one)
Use Cage = OFF; Smooth Normals = OFF
Size: 2048
Antialiasing: 1x (i.e.: none)
Closest Hit If Ray Fails = OFF (important! You want to see the failures so that you can optimize the distances; but even for the final thing you don't want it to take wild guesses... see below)
Tangent Space = ON
Filters: Checked OFF
Raycaster: Memory Conservative (It's faster than the fast one and doesn't crash)

The preview screen is too small, so just open the result in Gimp, and keep File -> Revert -ing at each iteration.
After optimizing the distances in binary search fashion, you'll still see a lot of spots where it either fails or picks the wrong surface.
So, in Gimp, use the Select Regions by Color tool, with a tolerance of about 30 or 40, and click on a flat blue area. This basically selects all the normals (texels) that have reasonable values, leaving out all the off the wall ones.
Now Select -> Inver, and bucket fill (Entire Selection) with 128,128,255; the default normalmap RGB.
With the selection still on, do a descending fibonacci radius Gaussian Blur multipass: 89, 55, 34, 21, 13, 8, 5, 3, 2, 1, 1.
Select -> None
Filter -> Noise -> Scatter RGB (uncorrelated, just 1)
Filter -> Map -> Normal -> Normalize Only.
DONE!

Image

Don't even try to feed the bumpmap to xNormal; not worth it; it would get chewed up in all the areas where rays fail. Better to do a bumpmap to normalmap conversion in Gimp, and blend them manually.
So just to show the normal corrections (no bumpmap here), first of all I show the normal corrections reversed, to exaggerate the artifacts of guraud shading:

Image

And with non-inverted corrections:

Image

I won'd deny that, besides corrections, there's also some artifacts introduced. This is because my fine mesh does have artifacts that subsurf introduced. So I'm going to apply the corrective normal map at half strength, by blending it with the normal converted bumpmap at 50-50.
I can't figure out how that multires-mesh feature works for LODs. Looks like you have to create the LOD first and then add resolution to get higher res. I doubt this will be very useful for making game-models. Maybe it can be used for baking highres details in the low version.
For what I've read, you can add a higher or lower poly res; but I'm not sure how it reduces poly count, and whether it allows you to define sacred edges. I imagine your hard edges would be respected.
Rivets:
As I recall I made a soft-brush for the rivets and increased the spacing. When you're satisfied with the spacing you can use this brush to trace over the paths you selected.
I think I'll keep doing them by hand; unless it can be persuaded to draw them on-grid, at the nearest grid point from the calculated curve and spacing, which I doubt.
z-buffer etc:
With the new node-system Blender can render multiple layers at the same time. This means you can render the z-buffer of the image to a gray-scale-image and presto you have a depth-map that can be copied onto your handcrafted bump-map.
Where and how do you do this?
Would be nice if we created a library of these details, so we could recycle them and get some uniformity across the models. In fact you suggested something along these lines in the Hornet-thread Chuck. I still have some samples I made there, awaiting wcu-dev crits.
I'm with you, and those examples you posted were gorgeous, if a bit too demanding in resolution. I think simpler ones would catch on. Say, for instance, like a recessed square box with like a lid in the middle, like a car's fuel point when the outer cover is open. Similar ones for like lube points. Some kind of shield generators. Grills, even though they make no sense in space, but heck... And weapon hardpoint stuff; handles for opening cockpits, hinges and locks for hatches. And last but not least, ailerons, like the one plate curves down under another. Greebles with a gazillion grooves need to remain full rez to look right, which makes them hard to scale to suit your needs.
And each should come with diffuse, specular and ambient occlusion counterparts, and a mask, so that it's easy to insert them.
I've just figured out another way to make this work. The new sculpt feature in Blender works with multi-res. It also supports custom-3d-brushes. If we use those pre-generated details as brushes you can literally stamp the details into your models. Then, because of multi-res, you should be able to bake the textures for the low-detail (the factual game-model) using a higher level of resolution.
Should work.
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Allright, here we go; first test of the new normal map; and it's not even 2k; just 1024. It includes baked normal corrections, so it's still the lowly, 5822 tri mesh, but it shades as smoothly as the 50k mesh:

Image

I know the grooves are too deep. I know the grills aren't there yet. Just a quick test. The important thing is that the rivets now look clean.

Image
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

I like that new bumpmap.

BTW: about huge textures. I've been working with 4k x 2k on the Hornet, and it runs well - on RenderMonkey, all by itself. I can imagine it would have a bad impact when used with other models, but I don't think it would be all that bad with modern GPUs, like that beast the 8800. And... for lowly cards, the engine can adapt.

BTW: I have it set to 1600x1200, 8x anisotropic, 6x antialiasing... on your 6600GT. Granted, the Radeon 9800 dies on that - so... you do have a point. But others also do have a point: in a little while, cards will handle pretty much anything. Remember that already they're experimenting on faster buses (those serial buses), so that memory bandwidth will also grow, along with computing power.

So... don't throw hi-res away... you're wise into trying to get decent results with the lowest resources, I always try that, but don't forget that you should also try to get wow results while eating up your hardware - when your hardware asks for mercy, that's where newer hardware will be comfortable.
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Okay, perhaps our experiences are too different. I was testing the Demon texture in the Vegastrike engine, in-game. Specifically, I was flying off New Detroit, and there were no ships within visible range. With diffuse and specular at 2k I had VERY choppy frame rate. I didn't even think it would be the textures, but I was wondering what was going on. It just occurred to me to try with 1k textures, and the framerate got a lot better. I was sure that 1k textures couldn't possibly be slowing down my (then) brand new 6600 GT, but just in case, I tried further reducing the textures to 512. Then the frame rate leaped up again, becoming really smooth.
I can't explain how you experience something different in Render Monkey; I don't have Render Monkey, and never used it.

Theories I can think of:
  • * Perhaps because in my case there were many other planet and ship textures loaded?
    * Perhaps your textures were compressed to .dds? Mine were PNG's.
    * Presumably there might have been other Demons in the New Detroit system which also gained texture size; would that have any impact?
In any case, like I said to Brad at wcplanet, if 512 looks too bad, I'll go to 1024. 512 is my goal; my target; NOT my religion. But we all have to have an idea of what we are aiming for, right? My main goal is to make a ship that looks good; but my second goal is to do so as efficiently as possible. If I could get the ship to look good at 256 I'd go for that; though I don't think I can. I just find it kind of absurd that people insist that I use a larger texture size, whether it's needed or not. As if making the texture as big as possible was the goal... Why don't they apply that philosophy to geometry? "Shoot for the stars! Go to 100k tri's!". It's like they think that polygons are a terrible thing to have, but pixels come for free; the more the better... What does it take to shake them off that mindset?

(Or, if the idea is to actually USE those 4 million texels times 5 textures, then exactly who is it that is going to put in the work to paint them? I'm spending enough hours on this as is. To target for a 2k texture, I'd probably would be working with 8k xcf's; but even working with 2k xcf's it takes minutes, sometimes, for a selective blur pass. And I get my share of Gimp crashes, like on Saturday I lost like 3 hours of work. I can't imagine working with 8k xcf's. But it's easy for someone not doing the work to say "shoot for the stars" or "we have standards for a reason"... Though nothing bothered me as much as "I agree with..." about that post... I mean the opportunism. But anyhow, maybe the idea was that I would go to a 2k texture and just fill it like a Persian rug with rectangles and lines quickly and mindlessly. But I've already said I don't like that style. If someone else wants that, they can do their own texture.)

I don't deny that there has been progress and there will be progress in terms of memory bandwidth. Sure. But nowhere nearly as much as the progress in processing power. What's that serial ram going to do? Double the speed? Good, but it's probably going to be slow progress again after that. Meanwhile, processing power will probably go up a 5 multiple.
Last edited by chuck_starchaser on Tue Mar 13, 2007 1:02 am, edited 1 time in total.
mkruer
Site Administrator
Site Administrator
Posts: 1089
Joined: Thu Jan 02, 2003 10:07 am
Contact:

Post by mkruer »

I think my exact words were “shoot for the moonâ€
I know you believe you understand what you think I said.
But I am not sure you realize that what you heard is not what I meant.

Wing Commander Universe Forum | Wiki
Wing Commander: The Wasteland Incident
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

[quote="mkruer"]I think my exact words were “shoot for the moonâ€
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

mkruer wrote:On the video cared front, you are correct the main issue is not texture fill rate but bandwidth. However there is another way to increase the bandwidth without increasing the speed, and that is the double the bus width, and this is exactly what Nvidia and ATI have done with this last generation of card. Nvidia has gone to a 320/384 bit buss and AMD is going to a 256/512 bit bus. My guess is this will continue, and chances are if the industry goes to multi GPU’s we will start to see a 1Kb bus in the next generation of card.(that’s 2 GPU’s each with their own 512 bit bus) This will allow the speed to be kept lower but at the same time keep the GPU’s fed.
Actually, you'll be seeing 1-bit busses. Or 4-bit... in the form of 4 1-bit busses, or things like that.
PCI-Express is a serial interconnect, it transmits one bit at a time, and doing that made the problem of transmittind data so much more easily tractable than they do with 1 bit what AGP did with 32 and years of hard-earned hacks.
Now that move is being tried on memory buses. The trick is that serial interconnects need bus arbitrators and, in general, a protocol (while parallel memory buses can be hardwired) - so it is actually a bit of a technical feat, but the resulting memory bus would be far faster (per wire) and far cheaper. The result is that you could pack... say... 32 serial buses instead of one 32-bit parallel bus, and get... I don't know... a bus 32 times faster?
Something like that.
It's a matter of theoretical limits - the parallel bus introduces complexities which limit bus design, and all those problems vanish if you use a serial bus. You just need a very fast bus arbitrator.
mkruer wrote:But we are digressing from the topic.
Yap.
chuck_starchaser wrote:GPU's *can* fill the entire screen with fully anisotropically filtered and antialiased texture texels. That day has arrived; but not in a game yet, where you can typically have a 3:1 overdraw. They can do so when you're filling the screen just once, as in Vista. But notice also the day has also arrived when GPU's can put a triangle into every pixel on the screen with full frame-rate. So my question still is, why are we always looking for any excuse to try and reduce poly count, yet always looking for any excuse at all to make textures bigger?
Let me answer that with one possible reason: because filtering algorithms and techniques are far more advanced in the texturing realm than on the geometry realm. FSAA still relies on brute force to get results, and that's quite inefficient. you can get much better quality if you play your cards right: avoid 1-pixel triangles, laverage filtering capabilities on the many texturing engines, and leave geometry for >1 pixel detail.
I agree though that you can have 3-pixel triangles by now ;)
If you're not wasting them, that is. But everyone knew that.
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

My god! Never been so terribly misunderstood. I didn't mean to suggest that we try to put a triangle into every pixel on the screen. Of course you wouldn't want to take the term FSAA literlally to the bank. All I was trying to do was to call attention to the fact that we are at a technology crosspoint, where the cost for a texel is within range of the cost for a triangle. And note I did NOT say "equal cost"; I said "within range" --'visual' range, if you press me to it :D

As for how many polys I want on the screen, I'd give the same answer as for textures, namely, whatever looks good, for a reasonable FPS cost. I don't like wasting triangles any more than I like wasting texels. I don't want to have the fragment shaders sitting idle waiting for the vertex shaders' output any more than I want to see the geometry pipeline doing nothing waiting for the shaders to be free.

But why did I want to draw attention to the triangle vs. texel cost range?

Just that a lot of people, Brad and MK being notable examples --and teaming up against me--, seem to believe that a triangle sells for a thousad texels. Am I exaggerating? Let's see: Brad says a ship should be 3k tris max, and MK wants to use 2k textures for them. 2k x 2k = 4M texels, x4 bytes = 16 megabytes, x (diffuse + specular + normal + glow + damage ) = 80 megabytes. Got distracted, never mind; 4 million texels no matter. So 4M/3k = 4096/3 = 1365.33 texels per goddam triangle. Which means these guys are rampantly out to lunch: The comparative cost might be in the range of 100 texels per triangle --IF that; probably more like 25 or 10.

And if the argument is about esthetics, I still say geometry looks a lot better than texture. So why, WHY, do I always have to argue for geometry against a pack of rabid dogs barking at me how evil geometry is; and then have to defend myself against the same pack pushing texture size up my nose? Why? WHY? ***WHY?*** do I have to spend all this precious time arguing, instead of working?

And WHY can't those same people do something for WCU, or mind their own goddam business, instead of hampering other people's work by taking up all of one's free time in useless argument? After all, I'm going to do things my way. And if I have something to learn, I will learn it from experience, or from convincing arguments; NOT from people telling me what's the goddam standard, like I give a rat's ass; or that "such and so is a professional and an expert and he says so, so you'd better listen", like Brad does so often. They are wasting their time, AND my time, because when so much error and falacy is written down with so much conviction it has to be fought down, for the record, and for the benefit of innocent visitors that might read that and be misled.

And it's such a waste of time and writing. Brad and MK, for the Nth time:
I'M GOING TO DO WHAT I WANT; NOT WHAT YOU WANT.
Get it? Now Get!
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Slow progress, and it's 3:40 AM; better catch some sleep..

Image
snow_Cat
Confed Special Operative
Confed Special Operative
Posts: 349
Joined: Thu Jan 05, 2006 12:43 am
Location: /stray/
Contact:

Post by snow_Cat »

^ - - ^ b Meow, Looking good.
chuck_starchaser wrote:And it's such a waste of time and writing. Brad and MK, for the Nth time:
I'M GOING TO DO WHAT I WANT; NOT WHAT YOU WANT.
Get it? Now Get!
^ . . ^ That's good because I want voxels to make a comeback. :p
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

snow_Cat wrote:That's good because I want voxels to make a comeback. :p
Hahaha, well, you'll be getting some voxels, actually, with the new engine; with paralax... A raised point on the bumpmap can occlude lower points behind; and shadow lower points, say, on the left --if the light is coming from the right. Bumpmapping with paralax is sort of a voxel engine in miniature.
In the meantime, I could try creatin a brush for Gimp ... in Blender3D :D
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Just a passing shot...

Image

The taper on ailerons doesn't look good, but I suspect it's the lack of bumpmap shadows; Klauss' shaders should fix that. Note the slight bumps around the outer rim at the back of the engines? Need to work on them; just a concept for now... I was struggling with the problem of inconsistency between shading and geometry, whereby a very smooth curved surface is cut by a plane, and a very polygonal curve shows, which disagrees with the smooth shading. So I thought, what about making the surface agree with the polygonality just towards the edge? So I added very tapered bumps at the edge.
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Okay, I'm done with the bumpmap for now.

Image

Bumpmap-minding ambient occlusion is next.
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Here's with some shininess modulation:

Image

Image

Textures are 512, and very plain.
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Alright; here's the really close-up shot that earlier caused such an uproar and all that searching for drastic solutions:

Image

It is supposed to look pixelated; as it IS a close-up shot. You'll never see the ship this close in-game. Just a test for how gracefully (or not) it pixelates.
I think it does so very gracefully, if I may say.
Never mind the brightness of reflections on the cockpit. I kept brightening its specular color and it was never enough, so I made it white and I'm too lazy to fix it right now. Too many Gimp steps to do, to fix just one thing.
Frankly, I can't even tell whether shininess modulation is working or not. I think so, but I don't know how Blender translates color to shininess, numerically speaking.
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

Gorgeous.
I think it's good enough. Especially the third shot counting bottom up - whatever fix you have on the last one would only look better on that shot - it's just that that shot seems to make optimum use of the textures' resolution.
You do have a hi-res version... don't you? (cuz I want to play with it :D )

If you have any pending additions to make, go ahead, but in my opinion it reached target quality levels - all it needs is some detail work maybe (labels, scratches, whatever you had planned) and perhaps some more fiddling with overal material texture (ie - get shininess modulation right - I know how to play with that, you could leave it to me - and stuff like that).
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Glad you think so, Klauss; I was beginning to feel that I should greeble it some more, but I didn't have too many ideas left, and I do like it as it is.

Indeed, inscriptions and scratches are what's left.

I haven't done any PRT's, though; and I don't have a height map that incudes the xNormal smoothing correction; --i.e.: just for the bumps; but I guess this will do for now.

Okay, I'm going to switch to finishing-off mode. Been looking for info for hours on how to have Blender read hardness from a texture. Pressing the Hard button is obviously not enough...

EDIT: Also have to finish the glow map; there's a bit of yellow light inside the arms at the front; and was thinking of a tiny amount of green light in the cockpit, to make it look like transparent without real transparency...

Two questions:

Did you say your shader automatically scales up specular color with shininess, for specular highlights? Because, like for the cockpit, I really had to crank up specularity for Blender to make visible highlights at high shininess.

Did you ever get around implementing fresnel? Frontal specular highlights look too bright in Blender.
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

I think you also have to play with Blender's equivalent of the "Blending mode" - set it to multiply. If you do that, black is zero hardness, white is whatever hardness you set up on the material. It's not like that by default, I know. I know I had to do weird stuff to get it right too once - I'll see if I find New Constantinople's blender file (where I did that), and I'll let you know (I'll probably take a screenshot).
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
chuck_starchaser
Elite
Elite
Posts: 8014
Joined: Fri Sep 05, 2003 4:03 am
Location: Montreal
Contact:

Post by chuck_starchaser »

Sorry, I was still editing my post, as usual; I added a couple of questions.
Actually, I tried multiply and it didn't help for me.
klauss
Elite
Elite
Posts: 7243
Joined: Mon Apr 18, 2005 2:40 pm
Location: LS87, Buenos Aires, República Argentina

Post by klauss »

A) No fresnel. But it's easy to add if you need it - just count on it.
B) No automatic scaling. I thought of adding it, but since the artist can pretty easily do it... why bother? It could be an option though... in fact, yap, it's really easy. Ha! Remember that lookup table? You can code the desired intensity scaling on that lookup table - violá

And the screenshot... I don't know if it'll help you from what you say, but since it worked for me...
Image
Oíd mortales, el grito sagrado...
Call me "Menes, lord of Cats"
Wing Commander Universe
Post Reply