Nosslaks stuff

[Nosslak]

That is not a problem, if they are two sep. models, you can attach one model to the other with no problems.

Great!

 

Although engine wise, two sep. models use a little bit more resources than one model, as they need to be two entitites, too. This isn't really a problem, because A: the feathers use a different texture, anyway, so they cannot be combined into a drawcall with the mask and thus get rendered in a sep. pass, anyway. And B : there won't be that many masks in a mission (maybe a dozend tops?) that havign another dozend entities is a problem.

 

And if all fails, we can still hack the engine and create multi-model entitites (in fact, we want to do this anyway. Entities carry a HUGE overhead, and if the entity consists of just a model, you have that huge overhead just to render a little model. That is why SEED can combine entities together by creating a mega-model. And we really want to use that with instancing, anyway.

Sounds like instancing would help a great deal here as well. Is there any progress being made on implementing that?

 

Before we had the source, models needed to be constructed to work with the engine in the best way. Now I think we should fix the engine problems, instead of placing the same (stupid) restrictions on every model.

Yeah, that does sound like the most reasonable solution.

 

Yeah, if you can please make one or two "single-feather" models, and maybe even a "half-feather" or "teared one". We can really use more "flinder" objects that get spawned when the player destroys things - makes the world feel more real than the indestructable stuff we usually have

Sure, that won't be a problem. At least not making a full feather or two, I'd need to find new textures to make ruined versions.

[Nosslak]

I managed to make a surprisingly good looking LOD-version with just 12 triangles and the same texture:

I managed to cut down a few more polys on the regular lowpoly taking it down to a whooping 154 polys.

[Springheel]

I found one more place where I could easily optimize a bit more, which saved an additional 12 polygons. So all in all I've saved 28 polys

 

This modelling by committee is fascinating. If you really enjoy spending your time cutting down polys, then more power to you. But don't be mislead into thinking that 28 non-shadowcasting polys actually makes a wit of difference to the engine.

[Nosslak]

This modelling by committee is fascinating. If you really enjoy spending your time cutting down polys, then more power to you. But don't be mislead into thinking that 28 non-shadowcasting polys actually makes a wit of difference to the engine.

Yeah, you're probably right but I think every modeler should learn to properly optimize their models. I don't think I can optimize it any more without sacrificing the quality or making it clip into the face though so I'm just going to stop right here.

[Tels]

This modelling by committee is fascinating. If you really enjoy spending your time cutting down polys, then more power to you. But don't be mislead into thinking that 28 non-shadowcasting polys actually makes a wit of difference to the engine.

 

No, it doesn't. But 154 polies vs. 300 for the mask does make a difference - esp. if you have a ballroom full of AI. And if that takes the modeller one hour, I think that is an hour well spent. I agree tho that wittling it down further is not economical.

 

@Nosslak: the low poly looks fantastic I bet from 5m away nobody will see a difference. Can't wait to try it in game.

[Baddcog]

 

@baddcogg: I don't think adding 300 shadow casting polies to an AI is a good idea, esp. because the shadow casted is VERY simple - it could be as much as 10 polies or so and nobody could tell it isn't accurately.

 

 

 

What I said was to include the shadow mesh mainly for a moveable mask. But either way, the author can decide when to have shadows or not.

If a shadow mesh is not included then the author has no choice, and the player can move an item with no shadows when every other model in the game has them.

 

I'm still confused. Right now I am making a second set of feathers so wouldn't it be a good idea to just make it a prefab so we won't need to load in two similar versions of the mask at the same time (however small the model may be)? I also feel that it'd be a lot easier to make the skins for it as I think you could just do it like this:

 

Feather1 and 2:

- Default (visible with shadows without collision)

- NoDraw (invisible without shadows or collision)

 

Mask:

- Mask 1 with shadows and collision

- Mask 2 with shadows and collision

 

Then the mapper could unflag what they don't want as you said, there wouldn't be a bunch of confusing skins/materials (confusing to me at the very least) and the mapper would be free to mix and match however the see fit.

 

 

You are absolutely right, I thought I had added those so that the mask wouldn't clip with the face but I tried simplifying it and there were essentially no difference. You really shouldn't make any adjustments to the mesh after you've baked your textures (the normalmap won't match the highpoly as well anymore) but I think I managed to make it similar enough so that won't matter. I am just saving 16 triangles so there's not that much of a difference.

 

 

I'm not sure people understand how little 10 polys are. A cube takes up 12 polys and a single-sided plane takes 2 polys. If I were to make up the different sides of the mask with 2 planes and then cut a foursided hole in each for the eyes that would already use 16 polys.

 

Skins are really the easiest way to go for mappers. 1 model, 20 choices of how it looks.

 

Basically you need an individual slot for each piece of the mesh.

 

mask

feather

collision (male)

collision (female) since those two have different shape)

shadow (male)

shadow (female), again , different shape, different shadow)

feather style 1

feather style 2

 

Some of them need to start off nodraw, but if you put nodraw one piece 1 and piece 3 (1 material slot), then when the skin changes nodraw to the female mask it changes it on pieces 1 and 3 (but #3 is the male collision model, you don't want mask on it).

 

By exporting only one piece with nodraw, and one piece with metal you give yourself 2 slots to change.

 

Authors just don't use the base model since it has weird textures, they only use the skins.

As opposed to items like a pot. They can use the standard model which has say clay texture, or the first skin which has crackec clay texture.

 

You could even make some simple textures that just say

don't Use #1 on them so authors see right away.

 

It's just one of those model things that really easy to do once you do it. But if you keep waiting for someone else to do it for you it remains 'confusing'

[Springheel]

No, it doesn't. But 154 polies vs. 300 for the mask does make a difference

 

Do you have evidence that even 1500 extra non-shadowcasting polys makes a noticable difference to a scene? Tests I've done in the past suggest otherwise.

[Tels]

Do you have evidence that even 1500 extra non-shadowcasting polys makes a noticable difference to a scene? Tests I've done in the past suggest otherwise.

 

No, unless you are on old hardware.

 

But it is the principle - models should be optimized as much as necessary - but not more. If every model we have wastes a few dozend or hundreds polies we'd end up with a pretty big performance hit for the same visuals.

 

And I don't quite get what the critism here is - Noss can spent his time as he sees fit.

[Springheel]

If there's a criticism, it's that a lot of free advice is offered to people making assets for the mod. Casual feedback is great. However, in some cases, I imagine this advice could be a little overwhelming (especially when it comes in the form of, "can you please change X and please make extra Y and learn to do Z"). This has already been brought up once in relation to sounds--people have avoided trying to make them because they assumed, based on free advice, that the standards of the community were much higher than they actually are.

 

Furthermore, this free advice is sometimes offered by people who know what they're talking about, but just as often by those who don't. How many mappers were told they should go through their maps and caulk all the surfaces the player couldn't see? How many may have abandoned maps or never started because it sounded like a lot of work? In the end, testing proved that it made no difference whatsoever.

[RPGista]

Damn, just noticed those little welding details on the metal, amazing attention! Is the whole model done with splines Nosslak? Working with (sp)lines and surfaces (nurbs) is the main reason why I cant get myself to learn subdivision programs, I wonder which ones you use to come up with all these good looking models?

 

@Springheel - in relation to what you say, welcoming casual contributions is also good because it becomes a repository from which someone with the rest of the knowledge can come and finish the job.

[Nosslak]

No, it doesn't. But 154 polies vs. 300 for the mask does make a difference - esp. if you have a ballroom full of AI. And if that takes the modeller one hour, I think that is an hour well spent. I agree tho that wittling it down further is not economical.

Actually, the mask itself is the only thing that I've optimized and I just removed 34 polys (about 18 percent). Optimizing away 18 percent of a mesh without degrading the visual quality is probably a pretty good exercise but to be honest the original mesh was pretty flawed and could need a little clean-up even though it might not matter much in-game.

 

@Nosslak: the low poly looks fantastic I bet from 5m away nobody will see a difference. Can't wait to try it in game.

Yeah, thanks! Many of the polys for the regular version were there to prevent clipping and as I couldn't spend all that many on this it will be offset a little bit from the face so there might be some popping if changed when the player is too close.

 

Damn, just noticed those little welding details on the metal, amazing attention! Is the whole model done with splines Nosslak? Working with (sp)lines and surfaces (nurbs) is the main reason why I cant get myself to learn subdivision programs, I wonder which ones you use to come up with all these good looking models?

Thanks! I use Blender, Photoshop, xNormal (generates normal- and AO maps) and ZBrush (only for the plague doctor mask so far) to make my models. I'm not really sure of the terminology outside of poly-modeling but for the design above with the LOD-mesh I used bezier curves which I think might be a kind of spline and the white surface was just regular old polygons with subdivision surfaces (smoothing). For most of my models I do use curves, on some models (like this) to convert to polys and bake to maps or to deform my geometry, but you can make good models without them if you want to. I've never used nurbs so I can't comment on that, though AFAIK it is basically the same as subdivision surfaces so I don't see any reason to use them.

 

Trying to learn modeling is hard (at least with Blender as your first program) but when I was 13-14 ish (6-7 years ago) I wanted to learn how to make my own games, which led me to making real-time graphics so I tried to find a suitable tool for modeling and eventually landed on Blenders site. I installed Blender and had a horrible time, until I learned to love it by sitting several hours every day with it until I felt comfortable with it (I was very mtivated). So yeah, learning a 3D software for the first time is hard work but (at least Blender) have gotten easier since then with a revamped interface and lots of good tutorials. Being able to model as good models (IMO) as I have done for this mod is hard, but so very, very rewarding that it makes it all worth it.

Edited January 29, 2012 by Nosslak

[Baddcog]

Furthermore, this free advice is sometimes offered by people who know what they're talking about, but just as often by those who don't. How many mappers were told they should go through their maps and caulk all the surfaces the player couldn't see? How many may have abandoned maps or never started because it sounded like a lot of work? In the end, testing proved that it made no difference whatsoever.

 

Well, I still say it's good practice to caulk unseen surfaces, and you can do it while mapping and get 90% of them with very little effort. Whether or not there are many gains to be seen also depends on other things. I garantee I could make an example where it would make a difference.

But as far as mapping goes it makes a cleaner map, makes it easier to see what is what, and it allows you to toggle unseen faces to see your way around the map in editor better.

 

Furthermore I doubt we've lost a single mapper because they were recommended to caulk unseen faces.

[Sotha]

Well, I still say it's good practice to caulk unseen surfaces, and you can do it while mapping and get 90% of them with very little effort. Whether or not there are many gains to be seen also depends on other things. I garantee I could make an example where it would make a difference.

But as far as mapping goes it makes a cleaner map, makes it easier to see what is what, and it allows you to toggle unseen faces to see your way around the map in editor better.

 

Furthermore I doubt we've lost a single mapper because they were recommended to caulk unseen faces.

 

100% agreement. It is probably not worth to caulk surfaces after the geometry has been created, but mappers should definately learn to involve caulking in the standard building process. Caulk things as you go and before you clone things. If you can reduce drawn surfaces with a simple workflow change, it will be very useful with minimal extra work.

[Johannes Burock]

I think the best way of caulking is:

 

If you create a brush, caulk it, get it in position an than texture it where it can be seen / or should isolate the void or another area . So you can be sure all sides which are "useless" are caulked.

And how Sotha said, also if you copy something, first caulk it and than going same way.

Unfortunatly i use this technique of caulking every brush first just since some months...and of course, sometimes i forget it .

[SiyahParsomen]

Come to think of it, one of the benefits of building geometry heavly with patches, you don't have worry about caulk. In my early mapping times, I was depending on it a lot, but now hell no. There are more efficient ways to improve performance.

[Springheel]

Whether or not there are many gains to be seen also depends on other things. I garantee I could make an example where it would make a difference.

 

I could make a situation where it makes a difference whether you have filter-stage textures vs decals too. But that doesn't mean that it matters in most cases or that mappers should be told to make their own textures.

 

Furthermore I doubt we've lost a single mapper because they were recommended to caulk unseen faces.

 

I can say from personal experience that I thought mapping was a lot harder than it turned out to be based on the way it was frequently described.

[Johannes Burock]

Come to think of it, one of the benefits of building geometry heavly with patches, you don't have worry about caulk. In my early mapping times, I was depending on it a lot, but now hell no. There are more efficient ways to improve performance.

 

Thats right, the massive patchwork makes this point often irrelevant for me .

[RPGista]

@ Nosslak - I know exactly what you mean, nice to know that you kept at it, it was worth it! The thing with subdivision is that it's good for creating complex shapes and aproximations, but only solid models and "water tight" surface models are actualy meant to be used for serious technical modeling (example:

), so its hard to spend time learning them. Anyways, have fun with the zombie, would you say humanoid shapes are easier to do in zbrush, instead of building them face by face?

[Nosslak]

@ Nosslak - I know exactly what you mean, nice to know that you kept at it, it was worth it! The thing with subdivision is that it's good for creating complex shapes and aproximations, but only solid models and "water tight" surface models are actualy meant to be used for serious technical modeling, so its hard to spend time learning them.

Alright, thanks for clarifying, but that's not at all what it's like working with curves in Blender. Here's a video showing blenders curves (old interface though, so it's a bit easier now):

http://cgcookie.com/blender/2009/05/20/modeling-with-curves/

You can also deform other meshes along curves and this is what I usually do to wrap complex patterns around my models.

 

Anyways, have fun with the zombie, would you say humanoid shapes are easier to do in zbrush, instead of building them face by face?

Thanks, I haven't used ZBrush all that much (I've done maybe 3 models with it, no ZSpheres yet) but I'd say that for me it's easiest to make a base-mesh in Blender without any real detail but with good topology (edge-flow) that follows the anatomy which makes it easy to add details and then sculpt muscles, bones and such in Zbrush. Here's the third and latest model I've made with Zbrush (and Blender for the base-mesh):

Sorry for the shitty render, it's just an old screenshot that were already up on photobucket.

 

I haven't thought about what I'd like to do after the current FM I'm working on. But now that you mention it, I'd love to do a masquerade FM. it would be great if you could put a mask on and the AI wouldn't attack you. Like in Hitman when you wear a disguise. Although that would be limited to the ball room. Caught any where else in the building and they get suspicious.

I forgot to ask earlier, but if you're going to make a masquerade FM when would you need the male mask?

[Sir Taffsalot]

My current FM is 100 miles away from complete. So please take your time. But I am very much interested in doing a masqerade FM so if you can eventually do a male mask I would be V happy

[RPGista]

That alien looks really good though, particularly for a first exercise. I really like the hands and feet, very well detailed and interesting/realistic anatomy.

[Nosslak]

That alien looks really good though, particularly for a first exercise. I really like the hands and feet, very well detailed and interesting/realistic anatomy.

Thanks! It's based on the mudokons from the Oddworld games.

 

I was a little bored yesterday and felt that I sure could use some kind of pedestal to present my models better on which resulted in this highpoly:

Took about 2-3 hours to model. I'm having some problems making suitably low poly lowpoly mesh, as it has a pretty distinct silhouette. Right now the lowpoly is at about 2300 and the shadow at maybe 900 triangles or so. Is this too much?

 

I know I should probably finish up the masks but they're almost done and I'm easily distracted. Julio is helping me with a zombie concept so I can't start that just yet either.

Edited January 30, 2012 by Nosslak

[Baddcog]

900 might be a bit heavy....would be good to trim down if you can but I wouldn't discount the model because of that.

 

But, it looks great, and most likely it's one of those one per scene objects. How are you doing the shadows for the legs? Triangluar? flatplanes? I'd think with flatplanes a really rough shadow (kindof mentally drawing it out here) could be as low as 320 tris. The a decent round for the top, probably under 600.

 

Maybe you could do an LOD which is nothing more than shadow mesh changes.

[Springheel]

That's more than double any of our current tables, so it would be good to reduce it a bit. Looks good though.

[Nosslak]

900 might be a bit heavy....would be good to trim down if you can but I wouldn't discount the model because of that.

 

But, it looks great, and most likely it's one of those one per scene objects. How are you doing the shadows for the legs? Triangluar? flatplanes? I'd think with flatplanes a really rough shadow (kindof mentally drawing it out here) could be as low as 320 tris. The a decent round for the top, probably under 600.

 

Maybe you could do an LOD which is nothing more than shadow mesh changes.

The LOD idea does sound pretty good in theory but I don't know how useful it'd be in practice as this table belongs inside where there are no really long drawdistances where it'd be suitable. The legs for the shadows are currently box-based but cutting them down to a being triangle-based would probably be good.

 

That's more than double any of our current tables, so it would be good to reduce it a bit. Looks good though.

Just to be clear that was the highpoly version (added a clarification in the original post too). I'm not sure where to reduce it as I've done that a bit already. I guess I could replace some of the ornaments on the legs (near the top) with planes as I am planning to do for the ornaments hanging down from the wooden circle under the table surface but otherwise I can't do much without worsening the quality. Here's how the lowpoly looks: