Hey guys, I'm looking at creating a normal mapped gun model for use in a first person shooter. I'm building this gun specifically to be used in the Source engine and I'd like the final version to include a diffuse, normal and specular map.

I'm hoping I can get a few opinions on the best way to create the normal maps for the weapon. The two ways I know of creating normal maps are to model a high poly mesh and project it's normals data onto a low-res base mesh to generate a normal map. Or to use an application which generates the normal map from a greyscale image, usually a modified version of the objects diffuse map. For that type of normals generation I've used Nvidia's Photoshop Normal Map Filter .

I know that the later method doesn't produce results as good as a 3d rendered projection off a high-res mesh. And you can also run into problems with upside down UV chunks in your texture sheet needing flipping and re-rendering.

I'm mainly wondering if it's worth the trouble of sub-d modeling a high-res mesh for the gun, as well as the low res mesh. Will the end result be worth all the extra effort, or should I just stick with the Photoshop filter method? Especially for a fairly inorganic object like this. At the moment I'm thinking I'll just use the 2d filter method, but I'd be very interested to hear anyone else's opinions and experiences.

Just to give you an idea of what I'll be making, the gun will be something like this in terms of real world scale, detail and the variety of shapes in its form. Thanks.

Modelling it in high-poly will look much more consistent and natural (easier to capture all of the small edges that catch light etc.)
Plus it's easy to texture since you have a solid base of normals rendered from the highpoly, and seams shouldn't be a problem since it's generated automatically rather than painted manually.
However it's probably faster to do it by making a normal map for a lowpoly gun by painting a greyscale bump map, but I don't think you can ever make a normal map for a weapon model look as convincing using the high poly technique.

Put it this way: I don't know of any professional game studios who would use the greyscale painted bump/normal map technique for making a gun model for a game these days.

If you want decent results it will take you longer to paint a good bumpmap than to just model it out. well maybe not your first time you ever model a highres mesh, but after a couple you should be way faster at just throwing together a sub-d mesh, instead of spending a week painting a super accurate and detailed hieghtmap, that will never look as good as simply modeling. If you just want to throw something quick together and you're not really concerned with quality then sure just paint your normals. If you jsut needed a small, low poly world model or something you could get away with this, but modeling something up close in your view that the player will see as a large portion oh his screen you HAVE to make a good highres first.

I spend most of my time at work making weapons. I would definitely model the highpoly - it gives you great normals that catch all the right angles, and a solid AO bake to work with for creating a texture.

Not building a highpoly model cripples your ability to make a good looking model quickly, imho. The time you spend painting a great normal map and painting in non-directional lighting will be as long, if not longer, than actually building the highpoly, and the result won't be as nice.

Like EQ says... why do you call subd modeling "extra" effort when it's an understatement to say that it's easier to model it out than going through the painstaking (wtf is this even a word? that's a crap word.) process of guesswork needed to produce anything even remotely similar in quality level in 2d.

And, if you think it's too hard to model that gun, model another one.

Painstaking -> "pains taking", as in, suffering for the amount of work you have to do. I think. I may have made that up just now, though.

yeah you gotta model the high if you want any decent looking results. All the games I have been working on I have been making the high poly model for all the guns i have done and weapons. Things like text and logos can be added in photoshop but really you will have better results modeling the high poly.

Take a look at The Kriss thread in Pimping and Previews... I showed someone the method i use. It might not be the best but it works great for me.

Thanks guys, I appreciate the input. I think you've convinced me to go for the high res method.

I am a bit surprised that some of you think painting normal maps by hand is comparatively difficult and time consuming to modeling a high and low res version of an object. I've been doing a lot of normal map generation from 2d at work lately, frankly a hell of a lot more than I'd like to be doing. I'm the first to admit it's a shitty substitute for the real deal in most cases, unfortunately I have no say in the matter for the project I'm on. However, I can usually create a normal and spec map for a main character or around 3 props per working day. In fact the only attractive things I can see in this method are that it's quicker and easier than generating normals from 3d. Admittedly, most of the stuff I've been doing won't be seen too close to the player camera and is also fairly subtly applied in game.

Of course it's a flawed process, good for bringing up prominent indentations and extrusions but as for getting nice smooth curves there's no chance, since the tech works from a 2d map none of your object's geometry contributes to the normal map you generate. However, doesn’t that sound ideal for something like a gun? Most of the geometry detail can be modeled into the base mesh, other than that it's just a few details like screws, scratches, and some fine line indentations that I could think of using a normal map to bring out. Not like a character where you've got things like hair, subtle muscle detail, clothing folds and texture to think about. That's what I was thinking approaching this, now I'll be going for the high res mesh method.

Thanks again for your opinions guys. I'm never afraid of putting in extra time and work for a better result, especially with personal projects, I just wanted to be sure the results would be worth it.

I'll give your method a shot too Jesse, I normally just have a key bound to turn nurbs subdivision on and off as I work.

Oh yeah, I always thought "painstaking" sounded like something you might do to a vampire.

The gun does have a lot of curved surfaces for being basically a box, so in this case I'd go with the flow and suggest trying a high poly of it.

That said I predominantly work using heighmaps to create my normal maps at work. It's not that I don't particularly want to model a high res and render that out, but I've never done it, and while the learning curve maybe fairly shallow I just don't have the damn time to put on my uphill skis (horray for tight schedules). Lucky I'm basically modelling boxes really...

Anyway, I have to smack Ghostscape upside the head for this little nugget:

"The time you spend painting a great normal map and painting in non-directional lighting will be as long, if not longer, than actually building the highpoly, and the result won't be as nice."

I made a script to do exactly that you plonker [img]/images/graemlins/smile.gif[/img] Takes all of 10 seconds to create non directional lighting from a bump map. Sometimes takes a few passes and requires some masking out, but it's probably still less time consuming than an AO render in Max. [img]/images/graemlins/wink.gif[/img]

Also, rendering out high detail NMs to a low-res image always ended up looking like crap for us when we were doing tyres for Saints Row - painting the treads in Photoshop and generating the normal maps from that worked out much nicer and took a lot less time - and you have ready made masks for your diffuse and spec that way. Just the other side of the coin for you WhiteEagle.

I'd still go with the high res for the gun though, unless you have a 256x256 map or something - and even I could manage that high res [img]/images/graemlins/wink.gif[/img]. You can do a detail pass for scratches, rivets and doberries in Photoshop later.

Model the high you lazy bitch...will be good practice for you.

pwn3d by gavku


But yea gotta agree modeling it will give you the better practice and you can learn the parts you can speed model/ cheat with with both normal mapping manually with high poly and additional programs such as crazybump or nvidia plugin.

I usually go for all the major hard forms first and do nec' picking details in crazybump/photoshop near the end.

[ QUOTE ]
Model the high you lazy bitch...will be good practice for you.

[/ QUOTE ]

In that case maybe I'll do it thrice. Only the best men model thrice for practice... [img]/images/graemlins/smirk.gif[/img]

I have some good Giraffe reference If you need it...