Duke Nukem 3D Modelling Project part 4

[The Commander]

Originally Posted by Broiler1985

How to do this?

Look up an advance build FAQ.

[Hellbound]

Originally Posted by Tea Monster

Just to let you know, if we had normal mapping and dynamic lighting, the models might just look a bit more like this:

Started to play with ZBrush?

[Tea Monster]

Hell Yeah! It's amazing .

I've already done the Cacodemon from Doom and decided to try it out on one of the Duke models.

[Plagman]

Are you making those models from scratch in Z-brush or using it to add detail to existing models to bake a displacement map?

[Tea Monster]

I'm creating my low-poly game mesh in Blender or Maya (I'm still more comfortable modeling in Blender). Then I pull the game model into Z-Brush and sculpt the out of it.

I'm still refining my pipeline, but I'm using a combo of Maya, Blender and Xn to bake off the high-poly normal maps for the game models.

This shows the general process and gives you an idea of sort of what it will look like in game. The skin is still WIP.

[Plagman]

Looks great. Z-brush definitely is a cool tool when it comes to high-poly modelling.
In this case, though, isn't the low-poly model a bit too low-poly?

[Spiker]

Tea Monster this model is fantastic, yet it has very little to do with Duke unless you plan to release some mod or enhancement pack, do you? I know you don't. However it would very interesting.

[Tea Monster]

Thanks everyone.
@ Plagman - Too low poly? Is there such a thing? Sorry, I'm kind of old school. I could put a bit of extra detail on board, but I think it works out pretty well.

@ Spiker - Dammit, you've figured me out! The Caco was put up as an example of what this will look like when its finished.

[Lightning Hunter]

Tea Monster, I hope that model is for the Doom coummunity! So many people create Doom models for other games while the Doom community itself is completely lacking good modelers... It needs some people like you.

[Tea Monster]

I'm going to be reworking one or two of my Duke models into next-gen models for the portfolio. I was hoping to be able to back-port them into the HRP when I'm finished. Are we still holding on the 3K for monsters poly count? I was thinking of doing a 3.5/4 K poly Level of Detail model and putting that back into the HRP.

Any thoughts?

[fast-1]

Is it easy modelling.

[Luigi]

Originally Posted by Tea Monster

I'm going to be reworking one or two of my Duke models into next-gen models for the portfolio. I was hoping to be able to back-port them into the HRP when I'm finished. Are we still holding on the 3K for monsters poly count? I was thinking of doing a 3.5/4 K poly Level of Detail model and putting that back into the HRP.

Any thoughts?

I say go for it, there is nothing to lose.

[DeeperThought]

Originally Posted by Tea Monster

I'm going to be reworking one or two of my Duke models into next-gen models for the portfolio. I was hoping to be able to back-port them into the HRP when I'm finished. Are we still holding on the 3K for monsters poly count? I was thinking of doing a 3.5/4 K poly Level of Detail model and putting that back into the HRP.

Any thoughts?

I think the 3K limit still makes sense, because lots of people have performance issues with the HRP as it is. Remember, Duke 3D is an old school game where there might be 50 monsters on screen at once.

On the other hand, releasing high poly versions in md3 format would be still useful for modders and individuals who want to use them.

[Spiker]

I'm not convinced by those performance issues because if someone has problems with HRP he should play the game in 8bit. I would do like this. There's no sense in downgrading because of that especially when technology is developing very quickly and computers are much faster and much cheaper than few years ago.

[Jinroh]

Yes, I agree, at least until EDuke or JFDuke get adjustble LOD. I play in 8-Bit mode, because my laptop lags every now and again and I don't mind a'tall.

[Tea Monster]

After seeing some of the new video of DNF, I was wondering about offering some of them as 'Classic Duke' replacements for the new monsters. With a bit of work (OK then, a LOT of work), much of what is in the HRP could be made compatible with a next-gen game. The poly count for the HRP would be really low in comparison with a modern game, but its an idea.

Back to DN3D, wasn't someone thinking of implementing a LOD system?

[The Commander]

Originally Posted by Tea Monster

Back to DN3D, wasn't someone thinking of implementing a LOD system?

Deeper Thoughts, Duke Plus has a LOD hack in it. It works quite nice to,

[DeeperThought]

Originally Posted by The Commander

Deeper Thoughts, Duke Plus has a LOD hack in it. It works quite nice to,

It was good in some situations (mainly the situation I coded it for) but in other situations it could actually lower the fps a little bit. Also, it depended upon having models that look enough like their sprite counterparts that it wouldn't be jarring when a model in the distance became a sprite or vice versa. But in fact it often was jarring -- especially in mods where there is only a model and the "sprite" is just a black box. In case you haven't figured it out I removed it for DP 1.7

[The Commander]

But it could work for the original HRP where as the counter parts had there sprites. But I never really noticed any slow downs with it on?

[Plagman]

Lower the FPS? What kind of logic were you using to pick whether to draw the sprite or the model? In what event?
Also, the same kind of system should work with a set of different LOD models.

[DeeperThought]

Originally Posted by Plagman

Lower the FPS? What kind of logic were you using to pick whether to draw the sprite or the model? In what event?
Also, the same kind of system should work with a set of different LOD models.

The method I used involved making every sprite check its distance to the player using the findplayer command in EVENT_GAME. Using findplayer like that caused lowered fps in maps with large numbers of sprites. In an early version I used the old ifpdistl command for distance checking, but it was unsatisfactory -- I wanted to use a formula that required getting an exact distance to the player. Here is the code:

 
state lodhack
ifvarand peractor3 1  // if bit 1 of peractor3 is set, that means that model rendering is OFF on THISACTOR
{
 findplayer xydist
 setvarvar tempb framerate
 ifvare tempb 0 setvar tempb 1
 divvarvar xydist tempb
 ifvarl xydist 512 // restore model rendering to the sprite
 {
  subvar peractor3 1
  setactor[THISACTOR].mdflags peractor3
 }
}
else
ifvarl framerate 30 // model rendering on THISACTOR is on and fps is low
{
 getactor[THISACTOR].statnum tempb // bad things happened when switching model rendering on regular actors
 setvar tempc 0
 ifvare tempb 0 setvar tempc 1
 ifvare tempb 6 setvar tempc 1
 ifvare tempc 1
 {
  setvar temp 0
  ifpdistg LODISTANCE
  {
   findplayer xydist
   setvarvar tempb framerate
   ifvare tempb 0 setvar tempb 1
   divvarvar xydist tempb
   ifvarg xydist 1024 setvar temp 1
  }
  ifvare temp 1 // disable model rendering on THISACTOR
  {
   getactor[THISACTOR].mdflags peractor3
   orvar peractor3 1
   setactor[THISACTOR].mdflags peractor3
  }
 }
}
ends

[Hunter_rus]

The "ifpdistl" and "ifpdistg" commands were broken till 31.05.2008.
In your case you don't have to use the "findplayer" anymore and can switch to "ifpdistl" and "ifpdistg".

Also the "findplayer" isn't as slow as the commands which involve map geometry calculations(like the "ifcansee" and "hitscan"). You can use the "getticks" command to find bottlenecks which need optimizations.

[Tea Monster]

Originally Posted by DeeperThought

I...Also, it depended upon having models that look enough like their sprite counterparts that it wouldn't be jarring when a model in the distance became a sprite or vice versa.

The problem is that LOD models would have to be constructed. A lot of people like Zilppuri are no longer active and I don't know if they would be willing to go back and re-jig their model for different LOD versions. I could do it for the Queen and the Cycloid (I'm going to be re-making the Cycloid when the Queen is done), but I don't know if its worth it for boss models.

The other problem is that LOD models usually have to be manually constructed and very carefully too. There are scripts that will automatically knock-out low poly versions of your mesh, but a script has no eye for detail and the results can be atrocious. Sometimes you have to re-rig and re-skin your new lower poly mesh as well, which is a real pain in the butt.

All the bitching aside, it sounds like a good idea for the game. Can you coding people get together and work out a system? Then we can decide on different LOD levels (how many levels, how far away the change occurs and how many polys for each level).

[Jinroh]

I'd say just for a quick test, import the models into Milkshape and then use the DirectX MD2 tool to adjust the LOD with the slider. Output a few different models and there you go. Good tests.

I played around with it for some of my .MD2 models and it actually does a pretty good job. I had some high poly count models I wanted to lower and they all looked fine by me when I used it. I could even spit out a couple of lower poly LOD models tonight, if it is requested.

Heck I mean, I wouldn't complain, they're far enough in the distance so who's really gonna notice that much?

[DeeperThought]

Originally Posted by Tea Monster

The other problem is that LOD models usually have to be manually constructed and very carefully too. There are scripts that will automatically knock-out low poly versions of your mesh, but a script has no eye for detail and the results can be atrocious. Sometimes you have to re-rig and re-skin your new lower poly mesh as well, which is a real pain in the butt.

If that's true, then there is virtually no possibility that the feature would be used, and so it would NOT be a good idea for the game to spend time coding it. We don't even have f*cking palmaps yet, despite the fact that it was coded quite some time ago. Personally, I would much rather have palmaps than LOD. Without palmaps we have to use those shitty tints to fake the different colors and it doesn't look right at all. If people won't even make palmaps, what makes you think that people would undertake the more difficult task of making low-poly versions of the models? A lot of the people who made them are gone now, as are the source files. Moreover, anyone who made the models before is probably a better modeller now and would want to make it better -- not just a low-poly version of the same model. (And let's face it, many of the most important models were done early on and could be done a LOT better; it hardly seems worthwhile to make low-poly versions of models that don't even look good in their high poly form). So essentially what you are suggesting would require all the models to be made over again from scratch. It would be nice if that happened, but I'm not holding my breath.

[Tea Monster]

@ Deeper Thought - Yeah, thats right as far as it goes. It would be nice to have it in place for future models though. I'd like to revisit a few of the Duke models myself, but I have no idea when I will be able to do it due to Real Life issues.

@ Jinroh - if you have Milkshape and it has this capability, would you mind trying it with one model, just so we can see how it works? If it does a serviceable job and it keeps the animations and uv maps intact, that would be great.

From what I've been told this dosen't work with Maya, but I'll investigate.

[Roma Loom]

HUD Shrinker revised, glow-map and fake-lightmap (for blast highlight) used.

[Tea Monster]

Wow, that looks very cool!

[Spiker]

Impressive. I want to see it in game!

[NightFright]

If any additional skins/defs are required for the Expander (meaning #2554 and #2558 pal2), please provide those as well. Otherwise, good to know someone is still doing something!

[Lightning Hunter]

Originally Posted by Roma Loom

HUD Shrinker revised, glow-map and fake-lightmap (for blast highlight) used.

Completely awesome work there!

Please tell me you are going to do a new Shotgun next.

[KillerBudgie]

Originally Posted by Lightning Hunter

Completely awesome work there!

Please tell me you are going to do a new Shotgun next.

Now that would be something to see, wouldn't it?

[Roma Loom]

HUD Expander, item shrinker, item expander and crystal ammo are in progress.

The shotgun is quite easy to make but nearly impossible to animate properly in game (unless Plagman and TX are secretely implementing event-based animation system). I guess that's the main reason Parkar avoids revising it.

[Hunter_rus]

What's the problem with the shotgun?

There is the "nobob" parameter available for years and it was never used.
hud { tile 2613 xadd 0.09 yadd 1.45 zadd -0.89 angadd -25 nobob }

Only known problem is interpolation doesn't work for HUD models.

[Roma Loom]

JonoF's site is down. Where can I find the latest documentation upon DEF language?
Ok. I don't get 2 things.
#1 - defining the reload anim for the shotty makes it play model anim completely random ingame. Try this code and see it for yourself

// Shotgun (2613)
model "highres/sprites/firstperson/2613_shotgun.md3" {
   scale 1.9
   skin { pal 0  file "highres/sprites/firstperson/2613_shotgun.png" }
   frame { name "2613" tile 2613 }
   frame { name "2613" tile 2614 }
   frame { name "2613" tile 2615 }
   frame { name "2616" tile 2616 }
//   frame { name "2617" tile 2617 }
//   frame { name "2618" tile 2618 }
//   frame { name "2619" tile 2619 }
   anim { frame0 "2617" frame1 "2619" fps 5 flags 0 }
   frame { name "2617" tile0 2617 tile1 2619 }
   hud { tile 2613 xadd 0.09 yadd 1.45 zadd -0.89 angadd -25 }
   hud { tile 2616 xadd 0.023 yadd 1.177 zadd -0.89 angadd -25 }
   hud { tile 2617 xadd -0.2 yadd 1.29 zadd -1.1 angadd -25 }
   hud { tile 2618 xadd -0.464 yadd 1.417 zadd -1.1 angadd -25 }
   hud { tile 2619 xadd -0.464 yadd 1.417 zadd -1.1 angadd -25 }
}

I make the first shot - it plays ok. I make the second shot it plays the loop incorrect.

#2 Once I replace "flags 0" with "flags 1" it plays the loop only once in the whole game, but not in a single action. Is it supposed to be like this?

[NightFright]

Please see to it that any pickups you recreate have ABSOLUTELY IDENTICAL alignment compared to current versions. I would appreciate if I could avoid going through all maphacks again because some models have different orientation.

[Hunter_rus]

DEF language:
http://wiki.eduke32.com/wiki/DEF_Language
Note: the documentation is incomplete and doesn't represent all supported syntax. The documentation on CON scripting is up to date(some description are missing though).

I though that there is no interpolation support for HUD models but I learned that model animation isn't supported too. It just replaces original tile with specified model which is defined in HUD structure.

Animation during reload:
Spoiler:

show spoiler

 0 2613
 1 2613
 3 2616
 4 2616
 5 2616
 6 2613
 7 2613
 7 2613
 8 2613
 9 2616
 9 2616
10 2616
11 2616
12 2616
13 2617
14 2617
15 2617
16 2618
17 2618
17 2618
18 2618
19 2618
20 2619
21 2619
22 2619
23 2619
24 2618
26 2618
26 2618
27 2618
28 2617
29 2617
29 2617

[Roma Loom]

Hmm... The currently released HUD shrinker uses "anim" operator quite succesfully, except for I have to use "flags 0" but that's ok for this case. Anyway there are cases when we really need event based animation system.. just like for the rocket launcher in Shadow Warrior I made few years ago but haven't released due to the impossibility of putting it into the game. You see the original rocket launcher in SW consist of several overlapping sprites which are animated independently, so trying to replace them with a single animated model was generally a huge failure. Also look at the reactor #1088 in Duke3D - it's hardcoded so I cannot assign models animation sequence upon its frames - it just won't animate the model. Also in some cases I need to play animation via textures cycling insted of model's frames... oh well...

[Roma Loom]

But on the other side I have finally completed the item shrinker model along with it's crystal ammo. Grab it here.

[Tea Monster]

That is great work Roma. Is anyone doing the shotty?