Shows how to make a basic collision although it's a bad collision model because of the amount of tris and what not but it shows how to get models, textures in and make a shader and apply it and save the upk file.
something's gone wrong there.. did you draw that map by hand then convert? Someone please correct me if I'm wrong, but isnt a major problem with converting a bump map that the converter doesnt know which way up the uvs are- so if the map is entirely converted that way and all your uvs aren't orniented 'upwards' you're going to run into problems
edit: so if you wanted to use it entirely for a character, would it not be better to apply the bump to the character and render off a cage normal map? if you were determined not to model a hi rez
I've been throwing this around #md, figured I'd throw it up here as well. :]
I've been using this as something to help me build a better high poly/low poly workflow.
Few more tweaks here 'n there:
earthquake/per- unfortunetly I have to disegree with You- in our company lead artist (we working on huge x360 project for ubisoft) usually killing dudes that making such nonsense wires- THIS IS A BOX! You can make such wires in case of vehicles (evrybody wants to ride a car that looking great or have a gun with evry crew) but NO ONE wants to look at the "awesome round" chest :P why? as I said- this is only the chest, it is better to optimize it and move that tris in to a car, weapon even character. Anyway- imagine that You have twenty boxes in one room...
working on this head right now - it's ingame in D3 engine
however I'm not really content how it's turning out. Something's wrong with it, but I can't figure out what is bothering me...
any suggestions? I've looked on it for long so I'm probably blind for the issues it's got.
it looked still alright with just the normalmap - so something of the texture seems to kill it....
per- and here is the point- xbox360 isn't such powerfull as I tought before I knew specification for xbox artists- why xgames are awesome? postprocessing etc etc but for example it has very small memory for textures and next thing- You are right about the balance- prop like a chest, box, barrel etc must have less polies becouse level designer maybe wants to put a lot of them in one place
And the last- poly not poly, thing about vertexes too, vertex UV calculation etc
In fact- less polies=faster=better :P About confusing me- I'm not new in gamedev and I know what are You talking about. Anyway ingame after removing that polies on that chest THERE WILL BE NO DIFFERENCE in final effect. Good realtime artists doesn't think- I have 1k poly for that so model up whole details! Good cg artist think- I have 1k for that... how can I make that 800 tri with same quality? We have normal maps, we have parallax, we have specular- USE THEM!
Cheer!
EDIT: and something like "insignificant optimization" doesn't exist
So you go on to say that texture memory is the problem, but you should optimise your triangles more than anything, and then use things like parallax which require extra texture memory to get back the quality? You're really going in circles contradicting yourself with that one. In almost every case adding 50 or 100 polys is going to be wayyyyy better than throwing an extra texture map in the mix, you should know that.
Really it cant be stressed how important *draw calls* are, much more important that the # of polys that can be drawn, so if you've got 5 boxes, or 10, the hit you're taking is from how many draw calls you get per box, not that drawing 5000 or 10000 tris. And the difference between 4000 or 5000, or 8000 to 10000 is not going to be noticable, at all. If you throw 50 boxes in the scene the hit you'll notice will be from drawing 50 different objects. You'll hit that bottleneck way before you hit poly limits with a prop this simple. Thats why it makes virtually no difference.
Mad: Your lead artist is wrong, sorry but he is. The mentality that drives the philosophy you described him as having ends up drawing more attention to the 1 and ignoring the whole.
Polycount is not the issue anymore, texture resolution / shader complexity is the issue on 360/PS3/PC.
On the surface, of course, optimizing everything is good sure but the optimization you demonstrated is half wrong because you detailed the need for taking out those horizontal edge loops which help ensure a nice smooth curve
on all axis.
It's fair enough to take out the middle edge loops on the
the corners all round, but you shouldn't state the need for the horizontal loops removal or the simplification of the handle because these are the key areas in implying the form is not just a box.
And with polycount, its more an issue of GPU vs CPU bound limits, you can optimize the polycount all day and it will reduce the load on one area but not the other.
In the end , in the case of this box chest, the best way to optimize it would be to ensure that each long and short side is mirrored so you can use double the texture res
and unwrap things into a fat L shape with the empty corner of the texture being saved for the handle.
In a case like this also it might actually be worth him leaving the middle edge loop in the corners for correct
mirroring on the texture in the mid point of the horizontal corners.
Thats the thing with normal maps, they are never going to render as sharply at the same res as the previous technology
so you really do need to use a little higher polycount to offset the damage from that and push much more optimized unwraps or reuse of textures across multiple assets.
Optimization is the key and the bread of all days, for meshes, uvws, maps, etc. In this case, that box needs optimization in my opinion. There are too many polygons wasted. I think this kind of model will never be seen in game so close, so it's not needed so much detail.
I've bought the newest games of this november (gears of war for pc one), and sadly.. i'm shocked how slow they run in my machine. I turned geometry to medium levels of detail and the fps were more aceptable but inappropiate. Shaders, texture details and postproduction were in the highest level. So, i'm not agree with that point. All counts.
One optimization here, here, here and there, can save too many fps in a game.
Mesh load is still a problem in games. We can't forget this. A whole level of a game is composed of multiple parts, if we ignore the optimization of each part because we "do nextgen", bad. All the steps in the creative process can be optimized, this inlude textures, uvw, shaders, etc.
Mirrored textures produces the butterfly effect. This should be used only in certain parts. This is something that should be avoided in next gen due to the bad aspect.
Not all ppl can play with the ultimate pc for gamers, a Nasa super computer as users commonly say.
With this said, if you can optimize, optimize, programmers will do their optimizations too. If not.. they surely will need to do some patches to increase improvement heh.
Blaizer, while I respect your work and talent, you are defending and perpetuating the incorrect idea that there is some very distinct connection between polycount and performance. "Less polys = better" is simply not the case, there are loads of other things to consider when talking about optimization. Lowering your polycount may not necessarily yield better framerates and in many cases you will end up sacrificing quality unnecessarily trying to reach some magic number.
You sound offended by what I said Blaizer, looking back over what I said to Mad, I can see your reason, I probably could have chosen my wording better.
The heart of what you are saying is 100% true, until we make optimization our no1 priority in general, we will collectively, never make perfectly optimized games.
Fair enough, but that point doesn't have as much to do with
my own point about the fact that a lot of those optimizations you end up doing, don't affect things like FPS.
Oh and I wasn't sugesting he unwrap the cube and created that horrible mirror Rorschach effect, I hate that!! That's
partly why I chose my nickname
I always advocate making one side and 1 length of the box
be unwrapped and then duplicated and rotated around 180 degrees.
I'm a little surprised to hear such "veterans" say something like this...
The real issue isn't even can the systems handle it, it's will you even see it. How big is that box, 1 foot wide? Less? If you put that box on a tv screen in the background of an environment it's going to be microscopic. There is no way you will ever notice the extra detail put into smoothing the sides or defining it to be "more than a box". At the size you will see it it will look like a perfectly straight sided box, period. You'd have to be far more dramatic in your shapes to actually have them read at in-game size.
Not to mention I have an extremely hard time believing that the optimizations mentioned would make it impossible to achieve the same normal mapped effect. Sure, you'd have to re-render the normal maps, but oh no, 5 seconds of your time has been wasted!
The only difficulty you'll ever face normal mapping with less edge detail (on this scale) would be if you were trying to portray those edges as large curves with details that run perpendicular to the curve. (for example if it was a 4 sided "cylinder" that was to be a hose with ridges going down its length).
Rounding hard edges are exactly what normal maps are SUPPOSED to do.
Sandbag: LODs man, if its not going to be close its going to be LOD'd anyway, so what you're saying is sort of moot. Theres no reason that he couldnt use that amount of detail on the highest level LOD model.
I dont think anyone here was every arguing that optimising is bad or useless, just that being extremely anal retentive about polygon usage is not the end-all-be-all to getting a game running smooth. Infact far from it.
EQ: Why the hell would you want to LOD a tiny box? That's probably just gonna reduce performance (either through dynamic LODing or extra memory required to load the LOD mesh *as well* as the main mesh).
Gotta say in this case I agree with Madman and Blaizer, it's a tiny box which is going to be maybe a foot or two wide in a game, the visual difference between 2 bevels on each edge and one will be nothing at all, and while everyone is just talking about rendering polys, what about collisions? It can really add up if you just have loads of polys everywhere without an optimised collision hull. Also for drawing decals onto the surfaces, it will be slower the more polys you have.
I think madman's point here wasn't "make everything look like a box cos normalmaps will do all the work", but in fact "in this specific case, optimising the mesh will improve performance (even if just a tiny amount when used on a large scale) WITH NO VISUAL IMPACT AT ALL."
Also I agree with Rorshach
Anyway this has pretty much derailed the thread now but I don't really have much to post. Have a picture of a blockout map.
Danke, Per! And good points, I'll see what I can do - it isn't anything final yet so I'll be back with updates. These two guys are from the Italian comic Alan Ford, which is for me and many of my countrymen an all time favorite:
[ QUOTE ]
dealing with poly pushers is a pain in the ass
"next gen hardware can handle it easily"
but do players miss this extra detail? are they willing to spend more cash on frequent upgrades to see a slightly rounder edge?
why not spend the extra horsepower on interactivity instead?
that's the difference between cg artists and video game artists. we do more with less.
[/ QUOTE ]Agreed, but why spend hours/days optimizing something when you can use that time to create more. It's not really about adding in more detail but taking out too much. Hump one object for a week to satisfy some personal goal the player won't appreciate or pump out 3-5 things they will.
Being a good game artist isn't just about making the most of every poly, but making the most with the time you have.
Replies
And then the black paint could have peeled a bit showing the original layer underneath.
[/ QUOTE ]
Peeled, yes, but smudged and blurred?
samurai helmet wip for the mini challenge over at gameatisans.org
www.artbyjessemoody.com under the tutorials page.
Shows how to make a basic collision although it's a bad collision model because of the amount of tris and what not but it shows how to get models, textures in and make a shader and apply it and save the upk file.
Now make a Tutorial "how to get shades in UT3" please!
made with Crazybump.
edit: so if you wanted to use it entirely for a character, would it not be better to apply the bump to the character and render off a cage normal map? if you were determined not to model a hi rez
I've been throwing this around #md, figured I'd throw it up here as well. :]
I've been using this as something to help me build a better high poly/low poly workflow.
Few more tweaks here 'n there:
Cheer!
however I'm not really content how it's turning out. Something's wrong with it, but I can't figure out what is bothering me...
any suggestions? I've looked on it for long so I'm probably blind for the issues it's got.
it looked still alright with just the normalmap - so something of the texture seems to kill it....
added the ropes
And the last- poly not poly, thing about vertexes too, vertex UV calculation etc
In fact- less polies=faster=better :P About confusing me- I'm not new in gamedev and I know what are You talking about. Anyway ingame after removing that polies on that chest THERE WILL BE NO DIFFERENCE in final effect. Good realtime artists doesn't think- I have 1k poly for that so model up whole details! Good cg artist think- I have 1k for that... how can I make that 800 tri with same quality? We have normal maps, we have parallax, we have specular- USE THEM!
Cheer!
EDIT: and something like "insignificant optimization" doesn't exist
Really it cant be stressed how important *draw calls* are, much more important that the # of polys that can be drawn, so if you've got 5 boxes, or 10, the hit you're taking is from how many draw calls you get per box, not that drawing 5000 or 10000 tris. And the difference between 4000 or 5000, or 8000 to 10000 is not going to be noticable, at all. If you throw 50 boxes in the scene the hit you'll notice will be from drawing 50 different objects. You'll hit that bottleneck way before you hit poly limits with a prop this simple. Thats why it makes virtually no difference.
Polycount is not the issue anymore, texture resolution / shader complexity is the issue on 360/PS3/PC.
On the surface, of course, optimizing everything is good sure but the optimization you demonstrated is half wrong because you detailed the need for taking out those horizontal edge loops which help ensure a nice smooth curve
on all axis.
It's fair enough to take out the middle edge loops on the
the corners all round, but you shouldn't state the need for the horizontal loops removal or the simplification of the handle because these are the key areas in implying the form is not just a box.
And with polycount, its more an issue of GPU vs CPU bound limits, you can optimize the polycount all day and it will reduce the load on one area but not the other.
In the end , in the case of this box chest, the best way to optimize it would be to ensure that each long and short side is mirrored so you can use double the texture res
and unwrap things into a fat L shape with the empty corner of the texture being saved for the handle.
In a case like this also it might actually be worth him leaving the middle edge loop in the corners for correct
mirroring on the texture in the mid point of the horizontal corners.
Thats the thing with normal maps, they are never going to render as sharply at the same res as the previous technology
so you really do need to use a little higher polycount to offset the damage from that and push much more optimized unwraps or reuse of textures across multiple assets.
r.
I've bought the newest games of this november (gears of war for pc one), and sadly.. i'm shocked how slow they run in my machine. I turned geometry to medium levels of detail and the fps were more aceptable but inappropiate. Shaders, texture details and postproduction were in the highest level. So, i'm not agree with that point. All counts.
One optimization here, here, here and there, can save too many fps in a game.
Mesh load is still a problem in games. We can't forget this. A whole level of a game is composed of multiple parts, if we ignore the optimization of each part because we "do nextgen", bad. All the steps in the creative process can be optimized, this inlude textures, uvw, shaders, etc.
Mirrored textures produces the butterfly effect. This should be used only in certain parts. This is something that should be avoided in next gen due to the bad aspect.
Not all ppl can play with the ultimate pc for gamers, a Nasa super computer as users commonly say.
With this said, if you can optimize, optimize, programmers will do their optimizations too. If not.. they surely will need to do some patches to increase improvement heh.
Nice "bang bang" EarthQuake
The heart of what you are saying is 100% true, until we make optimization our no1 priority in general, we will collectively, never make perfectly optimized games.
Fair enough, but that point doesn't have as much to do with
my own point about the fact that a lot of those optimizations you end up doing, don't affect things like FPS.
Oh and I wasn't sugesting he unwrap the cube and created that horrible mirror Rorschach effect, I hate that!! That's
partly why I chose my nickname
I always advocate making one side and 1 length of the box
be unwrapped and then duplicated and rotated around 180 degrees.
r.
The real issue isn't even can the systems handle it, it's will you even see it. How big is that box, 1 foot wide? Less? If you put that box on a tv screen in the background of an environment it's going to be microscopic. There is no way you will ever notice the extra detail put into smoothing the sides or defining it to be "more than a box". At the size you will see it it will look like a perfectly straight sided box, period. You'd have to be far more dramatic in your shapes to actually have them read at in-game size.
Not to mention I have an extremely hard time believing that the optimizations mentioned would make it impossible to achieve the same normal mapped effect. Sure, you'd have to re-render the normal maps, but oh no, 5 seconds of your time has been wasted!
The only difficulty you'll ever face normal mapping with less edge detail (on this scale) would be if you were trying to portray those edges as large curves with details that run perpendicular to the curve. (for example if it was a 4 sided "cylinder" that was to be a hose with ridges going down its length).
Rounding hard edges are exactly what normal maps are SUPPOSED to do.
I dont think anyone here was every arguing that optimising is bad or useless, just that being extremely anal retentive about polygon usage is not the end-all-be-all to getting a game running smooth. Infact far from it.
Gotta say in this case I agree with Madman and Blaizer, it's a tiny box which is going to be maybe a foot or two wide in a game, the visual difference between 2 bevels on each edge and one will be nothing at all, and while everyone is just talking about rendering polys, what about collisions? It can really add up if you just have loads of polys everywhere without an optimised collision hull. Also for drawing decals onto the surfaces, it will be slower the more polys you have.
I think madman's point here wasn't "make everything look like a box cos normalmaps will do all the work", but in fact "in this specific case, optimising the mesh will improve performance (even if just a tiny amount when used on a large scale) WITH NO VISUAL IMPACT AT ALL."
Also I agree with Rorshach
Anyway this has pretty much derailed the thread now but I don't really have much to post. Have a picture of a blockout map.
I started a topic here:
http://boards.polycount.net/showflat.php?Cat=0&Number=249534&an=0&page=0#Post249534
lamps and pillars are mine... from that little side project i'm working on...
Like indian_boy, i work on the UnrealEditor4, but, not the GoW version ... the Roboblitz one.
Screens below are done in the editor.
I have placed lights to see something but I'll rework the entirelightning when all details will be done and placed.
"next gen hardware can handle it easily"
but do players miss this extra detail? are they willing to spend more cash on frequent upgrades to see a slightly rounder edge?
why not spend the extra horsepower on interactivity instead?
that's the difference between cg artists and video game artists. we do more with less.
dealing with poly pushers is a pain in the ass
"next gen hardware can handle it easily"
but do players miss this extra detail? are they willing to spend more cash on frequent upgrades to see a slightly rounder edge?
why not spend the extra horsepower on interactivity instead?
that's the difference between cg artists and video game artists. we do more with less.
[/ QUOTE ]
http://boards.polycount.net/showflat.php?Cat=0&Number=249534&an=0&page=0#Post249534
Thanks for the import tut Jesse.
[/ QUOTE ]
hey no problem...glad it worked for ya.
More Zbrushing WIP.
Im going to chop these hands off and replace them with the good ones I made here:
At some stage this guy will get another thread of his own, or I'll resurrect the old concept thread.
haven't seen work from you in a while scoob. would like to see this guy finished up inside a decade though
dealing with poly pushers is a pain in the ass
"next gen hardware can handle it easily"
but do players miss this extra detail? are they willing to spend more cash on frequent upgrades to see a slightly rounder edge?
why not spend the extra horsepower on interactivity instead?
that's the difference between cg artists and video game artists. we do more with less.
[/ QUOTE ]Agreed, but why spend hours/days optimizing something when you can use that time to create more. It's not really about adding in more detail but taking out too much. Hump one object for a week to satisfy some personal goal the player won't appreciate or pump out 3-5 things they will.
Being a good game artist isn't just about making the most of every poly, but making the most with the time you have.
This diffuse will get baked down to the low & serve as a base for my final texture. So, yeah. less cancer, more w1n.
some fine work in this thread!
here's a big dude i'm working on at the moment.