So I started this project, and wanted to take it all the way through to a finished production and thought I would as for feed back as I went along.

Here is a link to catalog pictures so you can see the original:
http://www.rossiusa.com/2012catalog/?catalog_page=19

Here is the current state of things:










I'm pretty satisfied with the current state of the low poly model. I'm assuming that because a game engine is going to triangulate the model anyway, that in places where I can reduce poly count by using tries is an acceptable practice? What about n-gons? There is a couple of places, most specifically on the sights that using an n-gon would remove quite a few polys.

(I plan on doing an experiment and comparing normal map outcome on the cylinder with and without the indentation modeled.)

Other than that, any critiques before I move on to hi-poly for map baking?

No ngons. You could possibly remove the edges that splits the stock in half, since it doesn't seem to be changing the shape in any way? Although you might want to spend more time optimising other parts, since the low poly is quite low as it is. Mainly the UV's since these will affect the actual number of vertices that have to be drawn. Have a read of http://tech-artists.org/wiki/Beveling but don't go overboard worrying about optimisation. No point having an efficient model if it looks bad.

So I got distracted for a while but I'm back and began work on this again...

Thanks for the link Bek. I think that was probably one of the most informative info I ever gotten in regards to 3D.

So I went ahead and cleaned up the low-poly, beveled some edges, and made sure everything looked good in 1 smoothing group and took care of a couple topology issues. The only exception was the rail; I have no idea how to clean that up without adding way too much geometry, so I figured I just see how it goes.

Then I went ahead and cloned everything and fleshed out a high-poly model.

I also took a shot at UVs. I am mostly happy with how they turned out. A couple pieces could be improved, I think, but aside from chopping them into several sections, I don't really know what else to do with it.

(If there is a way to collapse set of images/text please let me know, I hate mile long posts.)

LOW POLY PICS





HIGH POLY PICS









UVs

If this is for a first person scenario then I reckon, in the UVs, reduce the size of the foregrip and make the cylinder parts larger. The ends of the chamber cylinder are too small and with the current UVs, when viewed from first person, will be quite blurry when textured.

As I have it now, everything has about the same texture resolution.

It would make sense to to upscale the textures closer to the camera for a 1st person scenario, but what about a "hybrid" situation, where it could be used in either 1st or 3rd person? Should it stay even, or have a "less dramatic" upscale?

Thanks.

Here is just a quick update.

I did a test bake for the normals and it mostly came out clean.

Resolution of 2048 with the previously posted UVs.

Should I re-size the UVs, giving more res to the complicated objects such as the cylinder and rail, while reducing res on simpler objects such as the trigger, lever, ect... Would that allow me to get a crisp normal map on the cylinder at a 1024 res, or is 2048 acceptable?

Also, as the model is quite low res as it is, and some of the edges are visibly jagged, I believe I could do 1 or 2 higher res LOD models.

Low Poly:


Low Poly With Normal Map:


And here is an animated gif comparing the changes:
https://docs.google.com/uc?export=do...0djdko2ZjJxcU0

I find this and the Max ram-player nice for comparing the mapped/unmapped versions, but it would be nice to compare the changes in real time at any angle. Any suggestions?

The results just got me excited. Low poly starting to look good.

Always looking for C&C though.

Did you change smoothing groups on the normal mapped one? Cause the low poly without the NM looks like it has lots of smoothing group errors, but with the normal map, not so much. Unless NM covers smoothing group stuff up that much.

No, the smoothing groups are all the same.

The way it works in max is, for the normal map to show up, everything has to be on the same smoothing group. So, every poly on each object is smoothing group 1. Then, when the Normal Map is baked, it is baked to a single smoothing group, in this case 1. (It can be any smoothing group, as long as it is the same all around.)

This generally causes some poor lighting, especially in corners. The sharper the curve the worse the lighting. The normal map, however, basically replaces this. Instead of basing the lighting normals by face (in reality it is by vertex, but the interface manipulates by face, can be quite annoying) it is based on each texture pixel. Any lighting errors then, would be a result of a problem on the high poly mesh.

That's how I understand it anyway. I could be wrong, but that is the conclusion I have come to based on a couple of failed experiments. This is officially my first "real" attempt at normal mapping a model.

A question I kinda need input on before I continue is if 2048 textures would be good for a production gaming model. I took a look at Fallout:NV rifle textures, and they seem to be 1024. But most of them appear to also be symmetrical. So to have unique texture of the same quality, it would need 2x the res.

So I redid the some of the UV unwrapping and balanced the resolution scale better and I hand painted textures in Mudbox. I ran into a few problems along the way, but finally found some solutions, so I threw it all in marmoset and this is what I got. What do you think, I am quite happy with it.

(The textures are 1024)

A bright sunny shot to show textures:


A night shot for atmosphere:

I think it's a good model and it's awesome that you've finished. I've got a few suggestions:

This depends on what the model will be used for/in, but put some emmissive on the green ironsights?

The wood looks pretty good, but the metal isn't as convincing. I suck at texturing so I can't tell you how to make it better, but if you browse through threads here you'll see some crazy good metal materials. At the moment yours looks too noisy, so could probably benefit from less in the diffuse channel, more work in the spec/gloss. If you can use a cubemap to add reflections too that might look cool.

Also, on the revolving bullet chamber, is that a UV seam ? Might make sense to put that on the other side, or even position it so that when the barrel rotates as the player fires, it eventually stops facing up, hidden by the picatinny-like rail.

The iron sights are fiber optic so technically they ought to be semi transparent; but because they are designed to catch the light, a soft glow map might actually add a nice bit of effect. I'll have to try that later.

I'm not really keen on texturing, but figured it would be a good idea to at least be moderate at it. I was kinda going with a slightly stylized realism, but it defiantly could use a little tweaking.

I didn't think you could use a cube map in Marmoset, maybe I'm wrong? It calculates reflection based on the specular map, and I used the alpha channel to mask out the wood. I did paint the textures in mudbox with reflection, so that may help the metal a bit too.

Yep, shoulda rotated that cylinder so the seam was hidden.

I'm not completely finished with this project but I'm taking a Portfolio Emphasis class at the community college and need ~5 pieces, so this one is on hold so I can get working on another project. I do plan on making a couple of lods as well as animating it.

Thanks for the feedback.