m32 - FPS Weapon Wrapup
polycounter lvl 18
Finally getting a chance to wrap this up and I'll need a bit of advice.
High Poly:
There are a few spots that I might need to go back and rework, but it should be enough to start with.
Low poly as it stands now.
1) The cylinder revolves, so making that area reasonably smooth adds an extra hundred polys (6 areas to round). Most of the 40mm rounds have a shiny bottom rim, so the ammo would be adding a highlight to draw the eye. So that area seems worth putting in the extra polys, but what about the far end of the cylinder? Rounding that adds another 300ish polys.
The selected edges are what I was turning on and off for the smoothness comparison above.
2) Is that area reasonably well handled for the low poly? The bottom of the barrel protrusion should have a few polys timmed, but I am seeing a few hundred polys that could be pulled out, not 1500.
Quick spin here.
3) Does 8.5k sound reasonable for this grenade launcher? When I started I thought low 7000s would be easy, but most everything is rounded on this thing. If you look at it at 1980x1200 (like it would be in UT3) it looks jagged pretty easily.
Input on anything you spot is appreciated, more or less learning as I go on this.
High Poly:
There are a few spots that I might need to go back and rework, but it should be enough to start with.
Low poly as it stands now.
1) The cylinder revolves, so making that area reasonably smooth adds an extra hundred polys (6 areas to round). Most of the 40mm rounds have a shiny bottom rim, so the ammo would be adding a highlight to draw the eye. So that area seems worth putting in the extra polys, but what about the far end of the cylinder? Rounding that adds another 300ish polys.
The selected edges are what I was turning on and off for the smoothness comparison above.
2) Is that area reasonably well handled for the low poly? The bottom of the barrel protrusion should have a few polys timmed, but I am seeing a few hundred polys that could be pulled out, not 1500.
Quick spin here.
3) Does 8.5k sound reasonable for this grenade launcher? When I started I thought low 7000s would be easy, but most everything is rounded on this thing. If you look at it at 1980x1200 (like it would be in UT3) it looks jagged pretty easily.
Input on anything you spot is appreciated, more or less learning as I go on this.
Replies
If you had a "This_is_the_model_I_see_other_players_with" version, that would probably be a different, even lower mesh anyhow. You also have a lot of detail in the stock and the handle that wouldn't be seen much in an FPS either.
You could easily get another 1500 if you cleaned all that stuff off.
One area you could easily cut back on is the fact that you modeled in the grooves onto the screw heads on the rear of the gun & sights. This is a good instance where a normal map is going to be able to fake this very well. You are staring almost directly head on to the screws, with a good high poly source, you'd have a very difficult time telling whether it's modeled or a normal map.
Take a good look at the model, and find out what a normal map will be able to do just as well as geometry, and also look at areas that will not likely be visible on screen much. You have a bit of detail on the lower back of the stock, will that ever be seen on screen? If so maybe only briefly during a draw animation or possibly a 'pistol whip' animation. So most of those details can be rough in the model, and the rest done with normal maps. You'd be surprised how much you can get away with on a normal map. Obviously being a first person model you want to get a little higher detail in the geometry, but like I said, there are many things that even in first person, you wouldn't be able to tell was fake or real.
Lookin really sharp though man. Looking forward to seeing this unwrapped and textured.
EDIT: I agree with Ott about that, however, then I suggest before you even lay out the UV's, you figure out how you plan for it to be animated. If it's something simple and basic where it comes up, you shoot, you reload, etc. Then you'll be able to do that kind of stuff easily, but if you want to have some bizarre animations going on, it would suck to not have faces there you then have to add, and have no UV space for it, and would also have to reproject some stuff if you did squeeze it into the UVs.
One area I noticed....that little lip sticking out from the end of the barrel.....I think the normal map would easily take care of this although the only thing you would notice is the silhouette would be gone
After ridding backfaces (if you want to take optimizing that far) and normal mapping the screws and the like, I think it'd be worth the chip in the budget, even if you have to sacrifice some more polys in less noticeable parts of the model - these tubes are the main attraction.
Looking really great so far - I can definitely envision this in an FPS.
yeah then you have to make a complete new model for the third person and the level of detail and having to use another texture ? why not letting it be there ? its not a couple hundred triangles that are going to make everything slowdown , its to showoff too dont forget, and he might want to be able to rotate the gun besides the fpv point.
On that point:
Flewda: Good to hear the modelling should work.
Thanks for the pointing out those screws, that is the kind of thing I don't have enough experience with. Looks like that would be a savings of 28 per screw, and I might have made some similar mistakes which could add up across the model.
I have been trying to think through the animation (and what I'd like to do and what I can do are probably pretty different), but the core of it is good portfolio stills. Even if the renders look okay a jarring polycount is hard to get past.
A pistol whip with 15 pounds of grenade launcher, ow
EricV: Thanks.
The lip would be a 50ish savings. I wanted to get some draft textures before deciding on that. I hoped it would pick up an interesting highlight, but as you say the normal map would do the same and you'd only see the barrel in profile at the top of a reload animation.
I'll go back over the stock, that is well out of view and I've been generous with bevels.
JohnnyRaptor:
The internals of it are a bit up in the air. They count for fewer polys then I expected, and there is a position in the reloading when you could be looking right down the pipe. I'll look into it.
pliang:
I don't know that means. Is that the weird stuff on the single smoothing group low poly, or something wrong with the high?
edwardE:
I've been back and forth on that one, good to know it sticks out to you also. I wondered if that would be softened when it's textured (instead of flat white), but you're right, those are the star of the show. They'll spin, they're right in front of you, etc.
Johny:
Exactly, I'd like to get some 3rd person views out of it. Showing off would be cool, but that's probably a pipedream.
I've wondered why the first person weapon wouldn't make a good highest LOD (assuming it is only a few hundred polys). It's likely still using the same size, and number, of maps that the highest 3rd person lod would.
... and posts are happening as I write this and I only had 20 minutes to post :poly105:
___
I guess I want to have my 'rounded enough for a FPV' and 'look okay in 3rd person' too. I'll approach it with the FPV in mind, with a dash of absurdly complex animation. I'll try not to be too fussed by the overall count, but make sure I don't get develop bad habits (like modeling in the 't' of the screws, likely the barrel lip, etc).
Thanks for the input.
Johny, no... think about it.
If there are 10 types of gun, each with a 1024x1024 first-person texture, you do NOT want to be loading 1024x1024 textures for the third-person versions, especially if you can stream in the first person textures. In fact it's probably more optimal to have a completely different texture for the 3rd-person weapon, at which point you might as well be baking down your firstperson's diffuse/spec anyway to a new layout. And TBH for this sort of weapon nobody will notice third-person mirroring, whereas they might notice first-person mirroring, so you would probably benefit from an optimised third-person model.
Either way, it's a solid set of meshes here, and I agree that you should spend a few extra polys rounding out the cylinders, since they really are filling the view.
This is only really beinificail *if* your engine supports streaming, otherwise you're loading a 1024 AND an extra 512, which is just plain retarded when you can easily reuse the same texture. Not to mention the sheer amount of production time to create and maintain 2 seperate models and textures. What if the texture needs to be overhauled? Its just piling on more and more work for *very* little(if any) gain.
Its quite easy to just make an LOD of your fpv model for third person, and even if your engine supports streaming you're better off doing this and just using a sized down texture for the tpv mesh, if you're worried about not having the FPV mesh and TPV mesh onscreen at the same time. What do you save, a few hundred polies and waste hours redoing uv layouts, baking redundant shit etc. NOT WORTH IT.
Anyway this is getting kinda offtopic and nitpicky, personally I think Kary is going well with this model, and his polygon count is about right so he doesn't need to worry about deleting the backfaces, I'm just saying you could do that without causing all the issues that Johny seems to think it would.
____
It came down to 7644, despite adding smoothness to the cylinder and keeping the potentially excess detail (butt and barrel) that I want to take a look at. The scope needs a 5th going over and I'll find more loose polys when UVing no doubt so that figure might come down. It sounds like a reasonable number though so I'm not going to obsess.
I'll get started on the UVs tomorrow, just have to do a few sketches to figure out how to do eyelids before I start them.
The gun doesn't do justice to the level, but it's interesting to see what it would look like.
A lot of this stuff has been theory to me till now, so it was a bit of slow going, and redoing. Results aren't spectacular, but so far I'm pretty happy. Freakily enough I found a bunch of polys in the UVing and it's down to 7,517 now (where it shall stay :poly131:).
Hopefully the layout isn't too bad. I haven't done anything that complex before and it got a bit cumbersome towards the end. I've got some stupid ratio decisions in there, but they're not mistakes I'll make twice.
I guess the next thing to do will be to paint out the normal map errors, get new cavity maps out of that, and then start the diffuse and spec. Am I missing anything?
Blown away by how fast xNormal is, 24 seconds for a 2048 of a 6.8 million poly model.
Thanks again for the documentation you did in your contest entry EQ, it was a huge help.
Looking forward to the diffuse and spec.
Also...
It would have been nice to get finished in the contest, but I started modelling late and it was harder then I expected. I managed to pick something centred around round forms for a gun contest... clever!
Hah that face is a happy accident, logical though:
It looks a lot more chipper on the right.
jogshy:
Thanks. I was able to get the simple AO from mental ray, and it is working in the older version too. Again huge thanks for the tremendous program.
MoP and Flewda:
Good to hear. I'm hoping the painting goes well, it looks reasonably simple, but there is a fair amount of it.
EQ:
I am actually following your PolycountWiki AO tutorial. I have a nice light exploded AO in there, but it has been completely smashed by the heavy 'low poly alone' AO that I dumped ontop. I'll fix the opacities and paint out the moving bits tonight.
LEViATHAN:
Good point. That band of shadow from the arm would look pretty silly in a spin
P.S - agree with leviathan.
It is, and mine are. I have taken some liberties in making a few bolts that should be matte black steel, but stayed somewhat true to the original. A friend, with a lot more 2d acumen, is doing an alternate skin and it's heavily worn and turning out quite well.
I am looking for my textures to not detract from the model, and I'm still experimenting with them. The spec map is really interesting to play with -- the mottled flash of light on the scope, and the streaky highlight on the barrel, is what I was hoping for, but I'm still experimenting with the matte paint.
I have a spin animation, UE screengrabs, and the maps on this page. I'm not sure how to put flvs inline here or I would have.
Thanks again for the help and encouragement on this. I will come back and tighten up the maps soon, but I want to get some environmental stuff nailed down. The environmental work around here lately has been inspiring