Anti-Tank Rifle (WIP)
polycounter lvl 3
Hello everybody.
Getting right down to it, I'm currently modeling out an Anti-Tank Rifle (mostly as a portfolio piece, but if it turns out well, maybe I could sell it on Cubebrush or something).
At first, I tried taking a bunch of different reference images at 20% Opacity, and overlaying them to generate a sort of "master shape" to work from.
As you can see, this did not work. (However, this method does give decent results with simpler items, such as swords and hammers)
So I went with Plan B: Pick a cool-looking reference of an Anti-Tank Rifle that already exists (A Steyr HS .50, in this case)
Sketching over the reference gave me a chance to break down the gun into separate parts, and should hopefully make modeling a lot easier.
Other than that, it's just been working in Maya to block out the basic forms.
At this point, I'm not too concerned about details, topology, or polycounts, but I'm still doing what I can to keep things clean as far as quads and edge-flow. Ideally, I'm planning on building a clean mesh that can be exported to ZBrush for finer details and normal projecting, followed by a topology cleaning pass back in Maya, before sending it off into Substance Painter for Texturing.
That's about as far as I've gotten at the moment (maybe about 4-5 hours in at this point, from the moment I started gathering Reference Imagery).
Comments/Suggestions/Critiques/Feedback are always appreciated,
~ArkOfDarkness
Getting right down to it, I'm currently modeling out an Anti-Tank Rifle (mostly as a portfolio piece, but if it turns out well, maybe I could sell it on Cubebrush or something).
At first, I tried taking a bunch of different reference images at 20% Opacity, and overlaying them to generate a sort of "master shape" to work from.
As you can see, this did not work. (However, this method does give decent results with simpler items, such as swords and hammers)
So I went with Plan B: Pick a cool-looking reference of an Anti-Tank Rifle that already exists (A Steyr HS .50, in this case)
Sketching over the reference gave me a chance to break down the gun into separate parts, and should hopefully make modeling a lot easier.
Other than that, it's just been working in Maya to block out the basic forms.
At this point, I'm not too concerned about details, topology, or polycounts, but I'm still doing what I can to keep things clean as far as quads and edge-flow. Ideally, I'm planning on building a clean mesh that can be exported to ZBrush for finer details and normal projecting, followed by a topology cleaning pass back in Maya, before sending it off into Substance Painter for Texturing.
That's about as far as I've gotten at the moment (maybe about 4-5 hours in at this point, from the moment I started gathering Reference Imagery).
Comments/Suggestions/Critiques/Feedback are always appreciated,
~ArkOfDarkness
Replies
Since it isn't UV'd at present, I'm planning on adding more polys for better sculpting with Edgeloops once in ZBrush. (This should work fine for most of the parts, such as the scope, barrel, stands, etc., but the compressor and main body give some issues, and need a little touch-up back in Maya before sculpting).
Once the extra resolution is in, I'm planning on exporting the base Subdivision back into Maya for UV'ing, then back into ZBrush for Projecting/Initial Map Baking, then back into Maya to clean out the extra polygons and hidden faces, and I'll take it into Substance Painter from there.
I'm sitting at about 8 - 10 hours so far, most likely. Not terrible, but definitely not the quickest turnaround.
I appreciate the feedback I've gotten so far, and look forward to hearing more suggestions in the future.
~ArkOfDarkness
Disclaimer: I haven't scaled the UV's to be consistently sized as of yet - I prefer to leave that until the end, once I've figured out what faces can be deleted, mirrored, etc.
And, while I was thinking of of putting this next slide in when I got around to Normal Baking, I figure that putting it in at that point may come a little late for sharing with anyone who's following along with this.
(Modeling for Normal Baking - A Personal Memoir)
You'd think that 5 years of University study specifically for 3D Modeling and Animation would cover some of the finer points of Normal Maps, but surprisingly, there's a lot that just never gets covered (some things you just have to learn for yourself, I guess).
Specifically, the fact that Normal Maps REALLY can't handle 90-degree angles, like, at all. Half the time, the end bake looks weird and icky, and the other half of the time, you end up with just a blank patch where the detail would have been. (Not the most fun thing to learn on your own after hours of modeling, sculpting, and baking, with a deadline sprinting up to meet you)
So the smart and sensible thing to do is to have angled slants and transitions, so that the normal map can pick up a nice ramp to display. (And, as shown above, an obtuse angle reads a heck of a lot better from most angles than a 90-degree angle)
I share this in the hope that it can save at least one poor soul from hours of frustration and deadline-induced panic attacks.
~ArkOfDarkness
It's obviously waaaaaay too much surface information to cram into the tiny little UV Square, and trying to fit everything in there while maintaining proper scale would really hurt some of the smaller components (trigger, scope dials, the little feet on the front stand legs, etc.).
But, if there's one thing I've learned over time, it's that there is nothing wrong with using multiple texture sets for a single mesh. (There is NO ADVANTAGE in cramming everything into a 4k sized texture square, when 2 or 3 1024x1024 squares can do the job just as well)
In the spirit of forward-thinking, and the fact that I may want to sell this model, use it for games, or use it in future projects, my plan is to keep the main parts as one texture set (Barrel, Trigger, Rails, Body, Magazine Slot, possibly the Grip and Stock), and place the remaining assets into a second texture set for interchangeable accessories (Scope, Bullet, Front Stand, Compensator, Foregrip).
QUICK EXAMPLE ON TEXTURE MAP SIZES:
As an example, create two completely blank files in Photoshop. One should be 1024x1024 Pixels, and the other should be 2048x2048.
Leaving them completely blank, save them out as JPEGs (or whichever image format you choose).
If you compare the file sizes, you should notice that the 2048x2048 texture is roughly 3 to 4 times LARGER than the 1024x1024.
It makes sense with basic math: Length x Width = Area, so increasing a 1024x1024 texture to 2048x2048 doesn't double your texture size, but actually QUADRUPLES it.
Using this logic/math/witchcraft, it could be said that a general rule of thumb is to prioritize splitting a model into multiple UV Sets over Increasing your texture size, as long as you can get away with 3 UV Sets or fewer. (If you find that you'd need 4 Sets or more for necessary detail, you'd be just as well off increasing the texture size, and having to load fewer texture files)
Comments/Critiques/Feedback are always welcome,
(especially if you notice anything wrong)
ArkOfDarkness
Details are projected, parts are split into 2 separate UV groups, finally hitting Substance Painter for Texturing.
The Base Materials are laid out, and I'm only just starting to add the extra detail (scratches, dust, fingerprints, etc.).
I'm figuring on a typical PBR setup (Color, Roughness, Metallic, Normals) in 1k and 2k sizes (thank you, non-destructive size scaling!), and if I'm able to sell it after all is said and done, I'll create a few additional skins to add on for free. (Probably like a clean sci-fi set, a blinged-out golden fancy set, and a glowing hot metal set)
Comments/Critiques/Feedback/Suggestions always welcome,
ArkOfDarkness
All together, there are 3 Texture Sets that make up the final product: the Gun Body, Gun Accessories, and Ammunition/Magazine Types, available in 1k and 2k texture sizes. Because the Ammunition sets are just the Magazine and Bullet, the texture size is half of what the rest of the textures are (EX: The Ammunition Textures in the 2K Set are only 1024x1024, and in the 1K set only 512x512)
Everything is also PBR, with each set using Normals, Base Color, Roughness, and Metalness.
Total Polycount for the Gun, Parts, 1 Magazine, and 1 Bullet is around 10,500 Tris.
(All Images pictured here are using 8 2K Maps for the entire gun, and 4 1K Maps for Magazine/Ammunition Type - however, there are also several shared Maps between Ammo Types)
ALL AMMUNITION TYPES use the same Metallic Map.
"Normal" or Ball Ammunition: 1K Normal, Base Color, Roughness, Metallic Texture Maps. (So only the Normals, Roughness, and Base Color are Unique)
ALL "SPECIAL" AMMUNITION TYPES USE THE SAME NORMAL AND ROUGHNESS MAPS
(The only "unique" texture used for each type is Base Color)
Sorry if this one is a bit long, but I'm super stoked over how everything turned out.
Comments/Critiques/Feedback would be greatly appreciated!
~ArkOfDarkness
Because you've got a lot more room to improve this piece. You've concluded far too early.
I do actually have some renders of the High Poly version in ZBrush, and I'm definitely all-ears for any suggestions to push things further.
It is probably worth mentioning that some areas of detail (specifically for screws and holes on certain parts, like the compensator) were added in Substance Painter during texturing.
~ArkOfDarkness
I get that you mashed up several different weapons early on in the concept stage, but it seems like you've done little to massage the design into something that flows well and is cohesive.
To be clear, I am critiquing the way this was visually designed.
The picatinny rail does not look like that in real life. There are other edges to it. Please reference and recreate a real picatinny rail system.
Your ammunition well just jams into the body of the gun, and you get a weird tangent with one of the barrel vents. That is an unnecessary tangent between shapes and elements.
Throghout your model, your hard edges are incredibly sharp on the high poly. This will make for bad normal maps, and later on, bad mip maps.
But you need to spend more time in the concept stage. There's a lot more that can be done to make this look good.
I would say you want to turn your bullets and such around so that you cannot tell as much that it is just a duplicate model...