I whooped this ship up in an instant..

[Duality]

And I took hours trying to tweak it as well.

http://edice.arvixe.com/images/stridenrot.gif

It may look similar to one of those ships/planes from either real life or a sci-fi series because of it's triangular stylish look to it.

Actually, the ship chassias is called Mugelos. More to come with this description soon.

Yeah I understand that this is a super low-poly ship, I have a hard time looking for places where to add more detail to but I will figure this out soon.

[BradMick]

i dont mean to sound rude....but are you and everyone else on this board afraid of using polys/tris? Theres a HUGE world of visual detail you folks could be doing if you'd just, up the poly/tri count to where the industry tends to run. Here's an example of where to place detail, and what kind of stuff to add.

This is an original ship design of mine, not for Vega Strike but for my own little universe. Hope it helps. Remember, you can go up to about (for small ships) 8k tris and be just fine. This ship lands at 6,368 tris. it needs textures yet, but it should help. oh, still gotta finish the landing gear yet too. it only took me about, 3 hours to model this puppy up to its current level of detail.










Brad Mick

edit: heres a few more quick model shots. fighters these are. The idea here is to show a more, dynamic design deal. for the larger ship, the tri count falls to 4,334 and it has full landing gear. the smaller ship falls to 2,662 tris. You can do a lot with a little. these are actually low poly models of two ships for a Wing Commander Film i'm working on, hope they help as well.

[Duality]

Well needless to say I'm not the type of person who would do high-poly mainly because of performance wise, but I pretty sure I do really like your detailed model here.

I will include some LOD but remember, I only smooth my models once.

[BradMick]

those aren't high poly. your average poly count for a modern game is 8-10k tris. if you can get under, great. but the model should look good.

Those models are, for the big transport looking deal, 6,368 triangles, the fighter with the landing gear, 4, 334 Triangles, and the smaller fighter is only 2, 662 triangles. Those are by NO means high poly. that's standard game res tri count.

[Duality]

And it does sound a bit similar to Siden. :)

[BradMick]

eh?

[jackS]

"High" polycount is something that should never be the limiting factor in one's model.

One should always use as many polygons as one feels are necessary to make the model look the way it should look.

This is of paramount importance, especially with respect to future-proofing the artwork for when better graphics hardware becomes available. Support for less powerful cards can be achieved via the LOD system.

To recap:
1. Aim high - use as many polygons as are needed to make a model you'll be happy with at any technology point in the reasonably near future.
2. Make LODs to cover the spectrum of polycounts (20K polygons at 10x10 pixels = absurd waste of resources, etc)
3. If the model you've created is too resource intensive for the current generation of graphics cards, then, because you did #2, we can just bump the top level mesh(es) off, put said mesh(es) in storage, and move all of the LODs up. Then, when graphics hardware advances, we'll just add the top end mesh(es) back in - which is MUCH easier than having to say "oh, we've had a major advance in graphics throughput - can you all go and build us new shiny top level meshes?"

With LODs and appropriate tuning of the models to whatever current hardware can support (taking into account the number of different models having to live in memory and be displayed at the same time) I see no reason whatsoever to limit the polycount of top level meshes except by whatever 'minimum' number is needed to satisfy artistic rather than performance requirements ('minimum' because it would be otherwise rude to neglect that there are performance considerations to eventually take into account).

There are places for low polygon art - but those are LODs, and small/fast objects which will likely be displayed on the screen in vast number.

[BradMick]

the level of detail i model at, is what current graphics technology is at. not all my models peak at the 8 to 10k tri range. the fact that someone would consider a model lower than 2000 tris to be an acceptable game model...let me put it this way. to my knowledge Freelancer (which is fairly old now) uses models int he realm of about 5k or so, maybe less. I'm well aware of LOD's and all that mess. and everything else. What i'm getting at is this. If Vegastrike is ever to get to a point where its artwork really evokes a sense of 'wow' then the modelers creating content are going to have to start working at the current levels of detail. 8-10k (or lower) is perfectly acceptable. the average is around 5-6k tris however. I don't need to be told about LOD's and all that. And if you think that 10k tris is high res, good god almighty folks you need to seriously update your knowledge of the 3D graphics world. When i do a high res model, the typical starting range is 40k polys (80k tris). And it steadily moves up from there. Modern graphics cards can handle models in the realm of 8k. Shit, if you're playing HL2, then you're already seeing what the added polys can do for you. DooM 3 is the same way as well.

[jackS]

Tangent:

DooM 3 is the same way as well.

Actually, last I checked the Doom 3 polycounts were generally < 4000, but the bump mapping did wonders for visual appearance. Read a couple of random articles about that some months back.


Back to the point:

I think you're slightly talking around my main point. Particular polycounts are irrelevant. Aiming for particular technology points is only somewhat relevant. Future proofing by using however many polygons you deem necessary to be artistically satisfied with the result is what is important. From that point, you then achieve performance by whatever compromises are necessary at the time, changing the compromises as time and resources change.

[jackS]

If Vegastrike is ever to get to a point where its artwork really evokes a sense of 'wow' then the modelers creating content are going to have to start working at the current levels of detail

Here I think you're horribly short-sighted. An appropriate approach requires working beyond the "current levels of detail", or more accurately without regard to the current level of detail. Top level models should be made to satisfy artistic vision, not engine requirements. The limitations of the engine are entirely the domain of compensatory modelling, of a continuum of imitations in the image of the desired product. Either graphics hardware will catch up to the top level by the time we release or it won't, or it may have long since passed the requirements for the top level because detail wasn't needed on that particular model - that's not really important. Spend polygons according to the needs of a particular model, not due to some imagined budget.

[BradMick]

well, by your approach i can tell you what will happen. nothing. why? clearly you're unaware of the amount of time and energy it takes to produce what your'e calling a 'top level' model. which i would guestimate (if you wanna go the level of DooM 3's 'top level' models) in the Millions of polys range. A helluva lot, thats how much. Months of modeling (if you're just doing it in your spare time) is what that entails And you know, its great if you're going to be doing Normal Map generation or something of that sort. Saying that Polygons is irrelevant is a bit...while i can understand it and even possibly fall in love with that idea, realize its very flawed. A poly budget for is perfectly acceptable, and nowhere near short sigthed at all. It creates a specific framework for a modeler to work within and get the maximum detail for the least amount of polys. Also, it keeps the objects at a level that will actually be playable and not have to have umpteen bazillion models created just to get it to a level that will play on current technology machines. I think i had this talk with hellcat once before. Creating content for 10 years down the road now is dumb. it just is. While its a grand and noble goal, most modelers aren't that good. They just aren't. Shit, i have trouble doing those uber high res models. They're not easy, its an endurance test from hell. (mind you, i don't think i'm some kind of insane uber modeler, but i am good) The reason there is a poly budget (in my mind) is so that you dont 1) intimidate the hell out of a prospective modeler and 2) keeps all content at a level of detail which won't result in diverging visual qaulities. Also, right now...there is no point to creating a top level model of however many polys you want. besides, most modelers invest their polys into stupid places anyway, like having forty bazillion sides on a tube, when all you need at most is 8 and a good smoothing group setup.

Fighters/Small ships 8 - 10k tris is good. 1k maps are fine here.

Capital ships (depending on how many will be on screen at one time) 10k - 30k tris. These could have 2k x 2k maps no problems, maybe up as high as 4k x 4k.

Those are the ways i work. I get every bit of detail i want too. I never sacrifice my artistic vision for saving polys. It's not hard. Either way, the modeler still has to build his low res model off of his high res model, which takes time. You can't just magically turn down the detail without it looking like ass (unless you're using nurbs...but nurbs suck, so why you'd want to do that...is beyond me). Anyway. I'm not talking around your point. I'm just saying i believe its incorrect.

[etheral walker]

And 1 more thing: you can have pretty good results with wings3D, but you take the hard way. Behind the fact the handling is more easy, you miss many powerfull functions. try blender. beginnings will be harder, but after some time praticing you will see it's more easy to obtain good results.

[jackS]

clearly you're unaware of the amount of time and energy it takes to produce what your'e calling a 'top level' model. which i would guestimate (if you wanna go the level of DooM 3's 'top level' models) in the Millions of polys range.

I am reminded of the common business saying "Better is the enemy of good enough" and likewise of the 'law' of diminishing returns.

If the requirements for satisfying one's artistic desires for a game model are so strenuous, perhaps one should consider withdrawing from working on objects requiring real-time rendering until such a time as one will not live in a state of constant disappointment.

I will stand by my statement that I believe the particular polycount for the top level of detail model to be unimportant - this does not mean that context has suddenly become unimportant. One is designing models to be rendered in real-time on a screen of not particularly gargantuan size. One is designing space-craft for a video game. One has certain expectations about what the users require to have presented to them in order for one to have convincingly represented objects on their screens.

One should model until one will be comfortable with others seeing a model up close and personal in-game. If one doesn't have the good sense to figure out that a point will arise when the additional detail that one is adding is superfluous and adds no additional value because one has already created a model which meets the goals and requirements of such a niche, then clearly that modeler will not be an efficient one, but that sounds like a personal problem to me - I'd like to think that, even if that point will fluxuate from model to model and modeler to modeler, that ranges of reasonable agreement can be reached for any particular case.

Having no particular polygon budget isn't some carte blanche approval of spendthrift techniques nor an invitation to obsessive-compulsive perfectionism. It's a philosophical stance that the comfort level of the modeler with what they are modeling (in the context that they are making a model for use by the Vega Strike engine) is more important than how many polygons it took them to reach that comfort level, because that's the how we're going to get art that isn't going to turn to vinegar before we get the chance to uncork it.

[chuck_starchaser]

Case in point: Duality's ship might in fact look good if he's trying to give it the flat-polygonal look of an F117 Stealth Fighter. Might not even want guraud shading with it...

[Duality]

And yes, lots of polies can add lots of wonderful eye-candy effects and it's also very costly to the speed of the computer and the graphic card.

My only main problem is, in order to do LODs, I must find a poly-reduction program for any program that can export to xmesh because doing it manually is probably a waste of time for me.

I will try to find it.

Best program to do it is in Blender.

[BradMick]

i'm not saying its not something i can do, and yes. it is a tedious and taxing ordeal. one can speak of things as if he knows all, but until he actually does something has no leg to stand on in my book.

the main point is this. design models for later on down the road later on down the road. do good work with what you know works in the present.

either way, things need to start looking a helluva lot better. so far, teh only things i've seen on this board that are anywhere near a 'holy crap thats awesome!' are from Stranglet, Ethereal, and when Howard was posting, him. Howard because he actually works in teh industry and knows whats going on with it (my mentor of sorts) Ethereal, well...hes just a damn good modeler. and strangelet, well...clearly that guy knows what he's doing and kicks ass at it.

sure, if you're going for the stealthy look...this works.

[charlieg]

I agree with BradMick and JackS. If a model is worth doing, it's worth doing well. You can't throw something together in 5 minutes and really expect a positive response. What's the purpose of the original model? It doesn't even qualify as concept art, it's barely even a draft of a sketch of a model. It's the type of model that could be the minimum level LOD of another [i.e. something to work with when the model is to be drawn in a couple of pixels].

At the same time it's difficult to be harsh because being harsh discourages prospective modellers. Everybody starts at a basic level, nobody turns into an Ethereal Walker or Strangelet overnight.

However, the project surely has passed the point where it needs models so desparately that pretty much anything can go in. The standards need to be set high for new models, and the contributors need to be aware of this. Only by having lofty aspirations can VS move forward graphically. The model at the top of this thread wouldn't have looked out of place in Elite I, surely Duality must recognise that Vega Strike is better than that. And in saying that, what did you [Duality] expect to gain from this thread?

[smbarbour]

I agree with BradMick and JackS. If a model is worth doing, it's worth doing well....And in saying that, what did you [Duality] expect to gain from this thread?

You answered the question yourself. The question that was implied is: "Is this model worth doing?" All models start somewhere. I doubt any of the modellers start by placing every vertex for the highest-quality model first and then dumb it down from there as necessary. By no means was he saying it was a finished model.

EDIT: Just as a question, is there anyone here that is willing to take models that others have done (with their permission) and improve them as well as create appropriate textures? (I'm not a great modeller (or probably even a decent one ), but I would be willing to rough out ideas that others could improve.) Are there people willing to do that or does everyone have the "If its not my creation, I'm not touching it." attitude?

[etheral walker]

My only main problem is, in order to do LODs, I must find a poly-reduction program for any program that can export to xmesh because doing it manually is probably a waste of time for me.

I don't think xmesh is the issue. You can use any prog which allow you to export to a format you modeler can load, I know some good and free ons, like melody, the normalmap generator from nvidia, which output normalmaps and lods. But the point is that automatic poly reduction will almost certainly look wierd. There is no better lods that manual ones;). In 3d the easiest way is the same as everywhere else in life: follow your master. For scifi there is many talented artists, I suggest you go to scifi-meshes.com, registration is free but needed, and browse wip forums to see the way the other follow to do things. Begin by reading tuts related to your package, and do all sorts of meshes, do not try to stick to one subject. Do some houses, peoples, animals, anything you can find a good tut for. It will help you to have a natural feeling with your package (ex: if you need to draw a sphere you should know IMMEDIATELY the keyboard shortcut, you should not disturb your ideas from the "artistic process" with handling questions)
When you begin to have a more intuitive feeling with your package you should begin to model what you want and find related tuts, even not related to your package

[BradMick]

for normal map creation, the best way to go really is Max. its pretty wicked. Do a render to texture deal, and all in one go you automatically get (if you set it up for this) the Spec, Color, Normal (using the bump map), and whatever else you want. All done right off the bat with no wasted effort. All you have to do is create your panel lines in photoshop, apply, and done.

[Duality]

Thanks for the tip.

I'll figure something out.

[etheral walker]

lightwave can do the same thing, there is a neet plugin for normalmapping
http://amber.rc.arizona.edu/lw/normalmaps.html
but yes, you have to move channels to bake the different maps

[BradMick]

well, thats more a plugin for the generation of normal maps off of high res geometry. the max deal just converts the bump to a normal in the baking process.

[etheral walker]

the approach of converting heightmap to normalmap can be done with photoshop and even the gimp( http://nifelheim.dyndns.org/~cocidius/normalmap/ ), but this way is less precise. My preferred way, if I want bumpmap conversions, is more like this one:
http://www.pinwire.com/article82.html
I kept the normalball.lws on my drive and use it as template for all my flat normalmaps
if you extract normalmap from heightmap you will have some problems with many sorts of parralax mapping, this one too (the most efficient for now)
http://graphics.cs.brown.edu/games/Stee ... index.html

[MamiyaOtaru]

Leaving aside for a minute questions of quality (I am glad to be seeing those addressed), that model reminds me very much of the Wing Commander Prophecy SWACS.