Animated grass - test map

[nbohr1more]

I'm more concerned about this mission:

 

 

 

http://forums.thedarkmod.com/topic/10003-so-what-are-you-working-on-right-now/page__view__findpost__p__204069

 

 

By the way, I have posted this to Moddb for some promotion:

 

http://www.moddb.com/mods/the-dark-mod/news/animated-grass-demo

[jaxa]

I didn't realize this was a public demo until I saw it on ModDB. Thanks, nb.

[nbohr1more]

The people love demos, give them what they love

 

Though, I (or somebody with a better PC ) need to make a Youtube of this for the spectators on the fence...

[Arcturus]

Heh, I was very skeptical... but wow.

Thanks Serpentine.

 

The Lion statues and fish were missing for me...

This is not good. Why some people don't see certain models? I'm affraid more people may have these problems.

 

I'm more concerned about this mission:

 

 

 

http://modetwo.net/d...post__p__204069

Please tell me that this is not going to remain "just a test map"

I practically abandoned this map... Last year I started completely new map. I've made a rough model of a castle, but it's far from finished. I didn't expoted it to Radiant. Then I started doing a photo realistic visualisation of a kitchen in Blender (which is also not finished), than the grass. Mapping is just too time consuming. I was able too finish this map only because I new how much was left and the fact that from the begining it was supposed to be only a small, test map was motivating. Right now I'm pretty burned out when it comes to mapping. Sorry.

 

By the way, I have posted this to Moddb for some promotion

Thanks!

[Arcturus]

The lily-pads seem(I havent checked the mtr) to be missing a normalmap in the ambient, _flat should take care of that.

Material seems to be fine.

[Springheel]

Here's a short Youtube vid of the highlights (recording was too dark underwater so I cut that out).

 

[Tels]

Heh, I was very skeptical... but wow.

 

This is some lovely work , and a really good showcase of some of the new features, as well as more advanced editing. I am already stealing a bit of this for a little something!

 

Re : Door texture, it's res is pretty solid already, I think it just needs the normalmap strength increased a bit The diffuse itself is perfect res.

 

You need a bigger monitor and then walk close to it - you can see the pixels

 

Arturus, that 2048x4096 door texture would be perfect for a TDM Highres pack!

 

(In any event it is still better than most of our 512x1024 or so textures for walls and doors Lovely

[rebb]

Beautiful work indeed

 

Average FPS on my old machine ( don't have Doom3 on my fast rig ) are around 6 FPS unfortunately, even at a very low "tdm_lod_bias" setting.

I guess this is due to the LOD problem mentioned in this thread ? Or is there another LOD CVar i am missing ?

[Arcturus]

On my computer changing LOD from highest to lowest gives up to 5 FPS more. Unfortunately animated MD5.mesh can't be hidden and it's the animated grass that needs most computing power in this map.

[Serpentine]

1920x1200 on a lovely calibrated 8bpp ips panel ;/

 

I think you're just looking at hard blending or resizing masks that have made the edges too sharp, while it'll be slightly better - you're also looking at 24mb just on a door normal.

 

A better resize on the diffuse(and maybe a little manual smoothing, I'm happy to do it) and baking at the native res would be the better imho. The bikeshed should be blue.

[rebb]

Oh i see, so its actually a mix of animated models and static SEED ones. Neat

I was confused at first that the tri-count wasn't significantly reducing when using something like "tdm_lod_bias 0.001", but it does seem to work as expected.

[Arcturus]

I use LOD but I didn't use SEED. I created the ground model and grass models and then I "seeded" them on the surface of ground using Blender's particle editor.

[nbohr1more]

That's fine other than your Entity count consumption (though the blades are illusory so it's not as bad as it looks). The real performance benefit would be adding "seed_combine" to the LOD grass.

[Arcturus]

After spawning the models via the particle editor I merged them. Animated grass are only 5 entities on the surface and one for the algae. Rocks at the bottom of the pool are one model (also spawned in Blender with particle editor and then merged). Number of static grass models had to be bigger. Otherwise it wouldn't be possible to use LOD and visportals effectively.

[Arcturus]

I uploaded a video too:

There's no sound, unfortunately.

[AluminumHaste]

I also uploaded a video lol, looks like everyone is getting in on the action. This one goes all the way up to 1920x1200:

[Berny]

WOW!!!! Beautiful - this looks so good!! I never want to see an outside area without animated) grass overall.

 

 

[Bikerdude]

Just played this on my system

 

GTX480

Core2Duo E6300@3.0Ghz

6GB DDR2-1066

 

All video settings in TDM to high and AAx0, ASx8 and res of 1680*1050

 

Lowest perf was 40fps, and got a weird fps hiccup (it drops to 10-15fps) when entering and leaving the water, which I'm guessing is linked the the VP that's in that spot or maybe for some reason the whole map is viewable hence the drop. But bloody awesome looking, this would be such a fantastic start point, I would love to make an inside for that castle facade.

 

@Arcturus, what say you, on another mapper building the rest of the castle, in keeping with the front it would need lots of organic patchwork inside...

[Arcturus]

Lowest perf was 40fps, and got a weird fps hiccup (it drops to 10-15fps) when entering and leaving the water, which I'm guessing is linked the the VP that's in that spot or maybe for some reason the whole map is viewable hence the drop.

When you go underwater the Visportal on the surface closes. However not entire world above the surface gets hidden at once. The FPS dropdown seems to be caused by mix of underwater overlay effect and portal sky (+ maybe the water ripple effect). When you go deeper remaining portals on the surface close, portalsky is no longer rendered and the FPS jumps high.

 

@Arcturus, what say you, on another mapper building the rest of the castle,

Why not. Of course it would be nice if the quality I was trying to achieve was preserved. Speedmapping is ok but I would like to see a map with high production value, so to speak.

 

in keeping with the front it would need lots of organic patchwork inside...

I was hoping I would inspire people to learn Blender, Max or Lightwave You can't make modern, good looking map without that, I'm affraid.

[demagogue]

That's just what I was about to ask reading through this thread... Looks like the punchline here is you can map quite a bit & exquisitely in a modeling program.

 

Arc, you should write up a little wiki tutorial about mapping in Blender and importing it. Not too much detail, but like the general plan, a rough idea of the pipeline involved, and some things that might work a little differently, visportaling, combining it with brushwork, performance considerations, lighting, gameplay in the area(?)... Or is there already a tutorial like that? Doesn't even have to be so much a tutorial, but if something is in the wiki then mappers can have an idea that the technique exists & it's something they can do and they have a general gameplan.

 

But also, if I were thinking about doing it, it'd be nice to see a short pros/cons or benefits/challenges list of doing it, and what situations it's good for.

 

@nbohr, you ought to put the DOWNLOAD link on the moddb page just above the screenshots, otherwise it's getting lost down below them, I think.

[Tels]

Rocks at the bottom of the pool are one model (also spawned in Blender with particle editor and then merged).

 

You know that you can achieve this with SEED dynamically? It even divides the generated models properly on visportals, it can cull them (so when you are inside the mansion, the rock models are not just invisible, but completely non-existing), and it can combine them etc. You could even generate the rocks dynamically so that their number scales with the "Object detail" setting in the menu...

 

I fear you have manually solved a problem that the engine could have solved for you.

[Tels]

That's just what I was about to ask reading through this thread... Looks like the punchline here is you can map quite a bit & exquisitely in a modeling program.

 

Er, yes, and it completely negates anything dynamic in the engine... E.g. instead of doing static things in DR (which is so yesterday), you do know static things in an external program. Sure, it might look better than some things (because you have more freedom), but it also is still static (don't let the "moving" grass fool you, by static I mean that you have always X grass models in the exact same spot), and it requires now another tool in the tool chain.

 

Or in other words, it is a step backwards (my personal view, of course).

 

 

(It's funny how much attention this map gets, I guess I really should have finished my pond testmap and released it public instead of just posting screenshots in the closed forum. Oh, and added animation of grass blades. Sad that this feature scores way higher than anything technical you code up over x months...)

 

Anyway, the result is still beautiful.

[Tels]

Just stumbled over this:

 

http://wiki.thedarkmod.com/index.php?title=Using_Blender_for_Doom_3_Modeling

 

Hasn't been updated for 3 years, tho, guess the same things get redisovered once in a while

[demagogue]

Er, yes, and it completely negates anything dynamic in the engine...

 

That's why I was asking for a general (& non-partisan) list of pros & cons. I don't pretend it's a knee-jerk solution for everything, but I imagined there's some things it does well and some things it gets in the way of, and it'd be nice of all of that were explicitly laid out.

 

For dynamic things I probably would keep it in-house, and I probably would use SEED for something like the grass. For certain static architectural elements like good looking crumbly walls or terrain or some byzantine grand foyer, working in a modeling program might be better. Especially if it's easy to mix DR stuff & imported models, I don't see why you couldn't use whichever tool is better for what you want to do in a hybrid way.

 

Edit: I'm especially conscious of the limits of patches; they only give you so many vertices to work with, when the engine itself can handle so many more, and don't let you do much with manipulating them or texturing anyway. So I generally think of using a modeling program (prob Lightwave for me) when I think of wanting to use a patch but doing more than it allows in DR.

[Arcturus]

You know that you can achieve this with SEED dynamically? It even divides the generated models properly on visportals, it can cull them (so when you are inside the mansion, the rock models are not just invisible, but completely non-existing), and it can combine them etc. You could even generate the rocks dynamically so that their number scales with the "Object detail" setting in the menu...

I fear you have manually solved a problem that the engine could have solved for you.

Rocks at the bottom of the pond are culled when you leave the water so I don't know what you mean.

 

Using vertex painting you can mix textures. Grass seamlessly blends with sand when the pool starts. There are three textures blending on the rocks. Three textures blending seamlessly on the wall of the castle. It's impossible to make such castle using only patches and brushes. Not only texture blending but the shape itself. Can you do such stairs with columns lions and hanging lamp using patches and brushes? It's also part mapping as I see it. I made this portal using only patches and brushes (stairs and lamp are models):

 

It took a month. I remember it as an endless struggle with limitations of patches and brushes. I'm saying that there are better tools for things like that. I don't say people should abandon patches and brushes completely.

 

Er, yes, and it completely negates anything dynamic in the engine... E.g. instead of doing static things in DR (which is so yesterday),

Half of the map was done using DR.

 

you do know static things in an external program. Sure, it might look better than some things (because you have more freedom),

It actually does look better because you have more freedom.

 

but it also is still static (don't let the "moving" grass fool you, by static I mean that you have always X grass models in the exact same spot), and it requires now another tool in the tool chain.

In the end it only matters what players feel. They don't care what tools mapper used.

 

(It's funny how much attention this map gets, I guess I really should have finished my pond testmap and released it public instead of just posting screenshots in the closed forum. Oh, and added animation of grass blades. Sad that this feature scores way higher than anything technical you code up over x months...)

You sound as if I have ruined years of your hard work when in fact I used it more extensively than anyone else (LOD).

 

Anyway, the result is still beautiful.

But it doesn't matter because I didn't use SEED.