the md3 vertex limit is higher than quake's mdl vertex limit, right?

Is there a quake vertex limit? 

http://www.speeds.quaddicted.com/dp000014.jpg
http://www.speeds.quaddicted.com/dp000009.jpg 

Do you walk up to people who are drawing in sketchbooks, or carving something out of a block of wood, and say that to them too?

Sometimes the pleasure in doing something the hard way is in the doing. 

I walked up to ppl who had said "I wish I could use the technology" and pointed them into the direction... 

i don't really use model props in quake much. if i want to do a map that would require detail models like that, then i do it for doom3. :P 

Now how can I do it in a good engine? 

Yeah, the md3 limits are 2 to 4 times higher than the quake ones on both triangles and vertices. The thing is though, unless you're making a really weird model with loads of breaks, then the number of vertices is always going to be lower than the number of triangles, but both have the same limit. So you don't usually run into problems, and if you do, the solution is just grouping more things on the skin, which is basically what taking a planar map is heading towards. 

what *are* the limits for quake .mdls? 

Ok, I made a slight mistake there. The standard limits, found in winquake and the released source are 2048 triangles, but only 1024 vertices. The compiled version of GLquake from iD actually has the triangle limit halved to 1024, but you can pretty much ignore that as just recompiling the GPL source gives you GLquake with this limit restored. There's no reason anyone should have trouble with that limit.

It's also interesting to note that unlike the max number of frames, which is 256 for networking reasons, you can increase these limits a long way without doing anything besides changing the headers.

In general I'd still say you're gonna hit the tri limit at about the same time as the vertex limit on most models. Once you've counted the vertices on the perimeter of a segment, a mesh of quads increases by 1 vertex for 2 triangles added(roughly speaking). So if you're unwrapping lots of small pieces the vertices are the rub, but for larger pieces the triangles will be more trouble. 

i had no idea the tris limit was so low o.o but thanks for the info! 

you shouldn't be using near that many anyway, because
a) it'll look totally out of place amongst the boxier stock models (and if you don't care about that, you should)
b) integerization will tend to destroy any fine geometry anyway. Remember the faces in Nehahra? Woof. 

my bones model took 1082 triangles/ 382 vertices.
And the mdl was large, 1150kb.

the granito model was 574 triangles/169 vertices and was 622kb. It is a huge monster I intended as end boss. The only problem I have it stops its attack in pain frames so one can easily kill it by constant shooting. 

Monsters' attacks are interruptable by default - you can cut off a hell knight's fire attack in mid-sweep for example. You just need to change your boss's pain function to cause pain every time (or nearly every time) and it should work, although that's if your .qc works like the original .qc, and I can't be sure that it does. 

Madfox, open up shambler.qc. The first thing you should look for is a series of functions; sham_pain1 to sham_pain6. These should look something like the pain functions of your boss monster, and I'll call them the "pain animation" functions. Now if you look at the spawn function for a shambler, right at the bottom of the file, notice the line:

self.th_pain = sham_pain;

What you should notice is that this is not the same function as sham_pain1. sham_pain is an additional function, which I'll call the "pain controller" function. It decides whether or not to bother sending the monster into pain. If you change that line to read sham_pain1(the first "pain ainmation" function), then every time the shambler takes ANY damage, it will run the pain animation, which I think is your problem with the boss. You might want to try this with a shambler, just to see it go wrong.

So now we need to see what the "pain controller" function sham_pain does differently.

void(entity attacker, float damage) sham_pain =
{
sound (self, CHAN_VOICE, "shambler/shurt2.wav", 1, ATTN_NORM);

if (self.health <= 0)
return;

if (random()*400 > damage)
return;

if (self.pain_finished > time)
return;

self.pain_finished = time + 2;
sham_pain1 ();
};

The english language version of this code is:
Play the pain sound.
If I'm dead, return.
Make a random number between 0 and 400, and if the damage taken is less than that, return.
If the current time is less than my pain_finished time, return.
Otherwise, set my pain_finished time to 2 seconds into the future, and play the animation starting with sham_pain1.

I'd say the two most important things the "pain controller" function does are:
1) Create a random chance for the pain animation to play or not (although many smaller monsters don't have this in their pain controllers)
2) Keeps track of the pain_finished variable - not playing the "pain animation" if this variable is greater than time, but also updating it to some time in the future every time a pain animation IS played. This is basically setting a minimum time between pain animations, and that minimum time should be long enough for an attack to occur.

So basically what you want to do is copy this "pain controller" function, rename it, and replace the call to sham_pain1 with granito_pain1, the first of your "pain animation" functions. Then make sure that th_pain is set to the name of the "pain controller" in the monster spawn function. Once it works you can tweak the values further. Hope that helps. 

sham_pain and wiiiine

I will justify this post (and turn the thread away from qc back to modelling) by saying I'm putting together my Maya export stuff for you all right now. 

Version 0.01 of lunmodelgen.exe and .py:

[ http://lunaran.com/files/lunmodelgen.zip ]
[ http://lunaran.com/files/lunmodelgen.txt ]

The .txt should explain everything, but only if you read everything in it. I also included an unfinished example monster! 

 

i found a useful utility that will deform a mesh into the shape of your UVs. you can use relax and pelt mapping now for quake models. :)

http://slidelondon.com/iv

when you use it on a mesh, it creates a new mesh with a morpher modifier on it so you can go between the unfolded mesh and the original. now you can just export the skin from with the unfolded mesh, then use the morpher to bring everything back together, slap a skin modifier on top and start animating! (if you use the original mesh and then try to combine it with the unfolded mesh, the vertex numbers will be screwed up) 

So I can finally have the skins painted properly without all the annoying black space outside. 

OK, stupid question (well, stupid person...);

How is the skin information in Quake different to normal texture/skin information in 3d apps?

I'm curious as to whether I can use the funky new 3d painting shiznat in PS CS3 Extended to skin quake monsters (or at least, a combination of PS and maya/lun's script) 

There are two types of UV vertices in a quake model, onseam and offseam. The offseam vertices are the ones not on the perimeter of any shape on the skin. These are exactly like UV coordinates in any other app: each such vertex on the skin is paired with one on the model, and every triangle which meets that vertex will have the corresponding triangle meet it on the skin.

The difficulty arises once you get to a vertex on the edge of a piece on the skin. You don't want all of the triangles which meet at that vertex on the model to join onto it at that UV point on the skin, because you would like to unfold them as separate pieces.

The quake mdl format allows such vertices to be designated "onseam". Then two copies of this vertex will be made on the UV map (but not the model) - one on the left hand side of the skin, and another half the skinwidth to the right of it. To determine which triangles connect to which copy of the vertex, each triangle in the model has a flag "facesfront" which is set by the model compiler. If the triangle faces front it is connected to the left hand vertex, otherwise it connects to the right. As you might guess, the original quake model compiler calculates this based on whether the face is front or back facing, which is why you get the front/back pairing on the original skins.

The alternative, which I support despite always listing it's negative effects, is to make none of the vertices "onseam", but instead make actual duplicates of those vertices in the model which are on the edge of a skin segment, so that they can be moved where you like on the skin. This allows for much better packing of the skin, since you can do things like mirroring halves, devoting more space to things which face front, all the tricks basically. 

Preach, that gave me more perceptive!
Point is I gave the granito.qc three pain sessions so
the self.th.pain = granito_pain_decide
I can change it but how do I calculate the other three pain session?

a closer watch:
http://members.home.nl/gimli/granito.qc 

It looks like most of what you need is there. First thing I would try is change all those pain finished things to equal time + 5, if it's meant to be a really powerful monster.

There's a small bug which means the first pain animation won't ever be called, you should change

if (r < 0.66)

to

else if (r < 0.66)

otherwise when r < 0.33 it's also < 0.66 so the second pain sequence overrides the first one.

To make the pain animations less likely to happen, you could change to something like

if(r < 0.05)
{ first sequence...}
else if(r < 0.1)
{ second sequence...}
else if(r < 0.15)
{third sequence ...}


That way if r > 0.15 none of them happen. You could even finish that with:

else
self.pain_finished = time + 0.5;

so that if it doesn't go into pain that time, don't even bother to check for the next half a second, which would defend it against pain from nails/lightning a bit more.

You may find putting all these things in might go too far the other way, I'm just throwing them at you so you have the options. 

That goes great!

First I copied the shambler's part of the sham_pain, but then I lost the other two pain sessions in the self.th.pain statement.

But indeed it has more power now. It can't be shot so easily and that's what bothered me in the first place for an end boss.

Hey, thanks for that splendid cham_pain explanation! 

wouldn't gl engines copy the vertex anyways for rendering?