Hey there. I really like the final look of your low baked mesh. I was wondering: do you re-model a low by boxes over your high, to make it fit for the bake? Or do you just reduce the mesh and remove all those edges straight from your high and detailled mesh?
Any plugins for Max you use to get the highpoly, like the turbosmooth or surface details? I usually get lost when doing such a gun.
I don't currently use any plugins but I do use built-in modifiers like symmetry, smooth, chamfer, etc. a lot. Max has a massive amount of built-in utility for modeling and it's worth exploring the documentation to fully learn the different tools and modifiers. In general I try to keep my environment non-dependent on third-party plugins. It makes it easier to move between different clients and workstations.
Hey there. I really like the final look of your low baked mesh. I was wondering: do you re-model a low by boxes over your high, to make it fit for the bake? Or do you just reduce the mesh and remove all those edges straight from your high and detailled mesh?
I make detailed blockout meshes that serve as the basis for both the highpoly and lowpoly. The idea is to get primary and secondary forms in the blockout so that the HP only involves preparing subdivision and adding tertiary details and the LP only involves optimizing and unwrapping.
Hmh. Interesting. I might have to change my workflow here. I always rush on having a good looking high mesh, and then reduce it, but it does not always perfectly match, so I have to tweak my normal etc.
The parts in red here are all one UV island, so those support loops help to control the shading and keep the whole thing from being a gradient on the normal map. Not technically necessary but it helps with texture downsizing and compression
The parts in red here are all one UV island, so those support loops help to control the shading and keep the whole thing from being a gradient on the normal map. Not technically necessary but it helps with texture downsizing and compression
All one UV island? But there are two different smoothing groups correct?
Are the materials made purely in photoshop or are you using a tool like quixel or substance? I'm curious for about your reasoning involving your choice whatever it may be.
I might also want to desaturate the brass on the barrel/gas tube a bit more.
Base materials looking great. Would be cool to see more unique wear to give it character, but I assume that will come after you've got base mats done for all the pieces.
excuse me, i just wonder how do you bake the normal map so clean and nicely and how do you render it with lighting in Marmoset? Can you share your setting ? if you don't mind
Are the materials made purely in photoshop or are you using a tool like quixel or substance? I'm curious for about your reasoning involving your choice whatever it may be.
I might also want to desaturate the brass on the barrel/gas tube a bit more.
are you using quixel or substance for the textures?
No, they're great tools but I'm not using them. My textures are still Photoshop fill layers, masks, patterns, and adjustment layers. I've been using PS for a long long time and I like to think I'm fast and efficient with it. I reuse materials and processes between textures, etc. I've used Substance a bit and it's nice. Mostly it's a better interface for a lot of things that are possible - but clunkier - in PS. Unfortunately I've found it to be crash-prone and hardware-intensive. When that changes or I get a new computer with a better GPU I'll move over to using it more. They're marketing it aggressively and the industry is widely adopting it.
excuse me, i just wonder how do you bake the normal map so clean and nicely and how do you render it with lighting in Marmoset? Can you share your setting ? if you don't mind
Most advice I could give is derived from things I've read over the years. If you want to get clean great bakes every time, it's important to understand everything from these sources:
Never adjust a cage by hand. Set up your lowpoly so it projects correctly instead.
Use control edges to fix skewing issues and overly aggressive shading issues.
One UV shell always equals one smoothing group. UV borders are hard edges.
Do one bake for one map. Don't combine different bakes to fix different problems. Learn to get it right in one shot.
Don't paint over bakes to fix problems. Fix the problems in the bake process itself.
About the Marmoset setup: it's just one point light, behind and above the camera, and inside the camera's transform hierarchy so it's always shining directly on the model from any point of view.
thank you so much for your advices because i just doing all wrong, i always quick bake the normal map by xNormal and always make mistakes just like you said. Hmm, look like i need to research before do anything. Btw thanks for your help about the diamond shape too, it's really cool trick and helpful and i did it
This looks amazing. What is the poly count of the weapon?
He already posted the gun at Artstation https://www.artstation.com/artwork/rOxEe There you can take a look at it inside marmoset viewer, On the bottom left, you can see the Polycount.
That bullet could use an edgeloop more on it's tip. Looking at the whole weapon, it looks a bit patchworky with all the different materials, is that on purpose?
I might get rid of some of the longer scratches on the flat, side portion of the receiver. They look too random to me honestly. Otherwise your texturing job is spot on dude!
Interesting project; I love AK's and it's awesome to see a unique twist on their design. From looking at the planned parts it seems only the barrels and handguards must share their type, while everything else can be mixed and matched - for example it doesn't seem that Barrel A can be used with Handguard B. Though, with an intermediate piece Barrel C could be used with Handguard B. The only other benefit I can think of having the barrel and handguard as unique pieces is down the road this will allow more barrel/handguard combinations, like if you were to make a shortened Barrel A, you wouldn't need a new handguard - otherwise I would've thought that the barrel and handguard could've been a single piece.
One UV shell always equals one smoothing group. UV borders are hard edges.
I prefer to think of it like this: A hard edge requires a UV seam, but a UV seam does not require a hard edge.
That's the most succinct way I can phrase it anyway. A hard edge requires a seam because to do otherwise would cause artifacts. A UV seam can have a hard edge for free because the vertices are already split (and it's usually beneficial to take advantage of that free shading change), but you're not required to do so. Smooth surfaces like a cylinder's side don't need that hard edge, so you can leave it smooth. It might not end up making much of a difference — but it's not the case that one UV island must "always" have hard edges at its boundary. But yeah, it works out that way 99% of the time anyway.
Also I've got a question and a comment about the magazine: I did one of these types just recently and I decided to have the small ribs as part of the normal map; you've added them as geometry. What are your thoughts on the pros/cons? I decided against it because I thought the extra detail there would be unnecessary and because it would make the UV mapping + LODs somewhat more fiddly. As for the comment, the size of the 7.62x39 round makes it look like the magazine is only a single-stack, perhaps because having two smaller rounds wouldn't sufficiently block vision of the mag interior..?
Another thought, did you have any trouble fully utilising UV space with some of the smaller modular pieces (like the grip or magazine)?
That bullet could use an edgeloop more on it's tip. Looking at the whole weapon, it looks a bit patchworky with all the different materials, is that on purpose?
I didn't want them all to look like they were made to go together out of the exact same materials, but I might unify some of the plastic parts later, not sure yet
I might get rid of some of the longer scratches on the flat, side portion of the receiver. They look too random to me honestly. Otherwise your texturing job is spot on dude!
Yeah it's a good point and I think many of these textures will benefit from a second round of polish.
From looking at the planned parts it seems only the barrels and handguards must share their type, while everything else can be mixed and matched - for example it doesn't seem that Barrel A can be used with Handguard B. Though, with an intermediate piece Barrel C could be used with Handguard B. The only other benefit I can think of having the barrel and handguard as unique pieces is down the road this will allow more barrel/handguard combinations, like if you were to make a shortened Barrel A, you wouldn't need a new handguard - otherwise I would've thought that the barrel and handguard could've been a single piece.
I don't mind altering the exact lengths of different barrels and handguards if needed to make them compatible. I think ultimately Rec C and Stock C will be the only parts that "have" to go together because of their unique junction incompatible with the other receivers / stocks (maybe I just won't make them).
I prefer to think of it like this: A hard edge requires a UV seam, but a UV seam does not require a hard edge.
That's the most succinct way I can phrase it anyway. A hard edge requires a seam because to do otherwise would cause artifacts. A UV seam can have a hard edge for free because the vertices are already split (and it's usually beneficial to take advantage of that free shading change), but you're not required to do so. Smooth surfaces like a cylinder's side don't need that hard edge, so you can leave it smooth. It might not end up making much of a difference — but it's not the case that one UV island must "always" have hard edges at its boundary. But yeah, it works out that way 99% of the time anyway.
Sure. i like UV shells = hard edges because your results will always be correct anyway and it's easy to automate "unwrap-by-smoothing" or "set-smoothing-by-UV-shells" to ensure consistency
Also I've got a question and a comment about the magazine: I did one of these types just recently and I decided to have the small ribs as part of the normal map; you've added them as geometry. What are your thoughts on the pros/cons? I decided against it because I thought the extra detail there would be unnecessary and because it would make the UV mapping + LODs somewhat more fiddly.
They'd probably look fine normal mapped, I just left them in because at the time it seemed like the physical depth would be nice to have. The whole side of the mag is one big relaxed UV shell so it didn't really affect that process, and LODs would just be collapsing loops. In general the mag LP isn't very well optimized, I might go back and reduce it. I probably made it too thin also, should probably be flared out to look more capable of holding 30rnd.
Another thought, did you have any trouble fully utilising UV space with some of the smaller modular pieces (like the grip or magazine)?
Not really, but since they're all textured at 2k the texel density for different parts is all over the place, like the grip is many times more dense than the receiver for example. For a real production many of the less essential parts like the grip or barrel could be downscaled so it kind of works out though.
I think ultimately Rec C and Stock C will be
the only parts that "have" to go together because of their unique
junction incompatible with the other receivers / stocks (maybe I just
won't make them).
Maybe Stock C could just have its attachment point changed to be adaptable with the other receivers (not marging with the dust cover), it would only have to be shortened slightly from the concept.
I like that you are texturing all of the parts separately, it'll make each piece more unique, and less uniform, but i feel like doing a final texture pass of the entire model all at once might unify the textures a bit, which it could use a bit of. You probably want the gun to feel like it's been assembled and used together for a while, and not something just slapped together today.
I prefer to think of it like this: A hard edge requires a UV seam, but a UV seam does not require a hard edge.
That's the most succinct way I can phrase it anyway. A hard edge requires a seam because to do otherwise would cause artifacts. A UV seam can have a hard edge for free because the vertices are already split (and it's usually beneficial to take advantage of that free shading change), but you're not required to do so. Smooth surfaces like a cylinder's side don't need that hard edge, so you can leave it smooth. It might not end up making much of a difference — but it's not the case that one UV island must "always" have hard edges at its boundary. But yeah, it works out that way 99% of the time anyway.
Sure. i like UV shells = hard edges because your results will always be correct anyway and it's easy to automate "unwrap-by-smoothing" or "set-smoothing-by-UV-shells" to ensure consistency
I totally agree with Bek on that Point.
Earlier in this thread you posted this picture saying all the is red, is one uv shell. So your shading without an normal map and without the geometry you added should look like the one on the right: Now your UV Layout for that piece should also look like the one on the right in the following picture: On the left side you can see how i would do it. I think that also the way that Bek tried to say, he would do it.
No if we bake both of this examples with the same highpoly, we should get the same results. And we get nearly the same results. But when you take a look at the normal map, you can see that the left side, where every hard edge is a UV shell, is way cleaner.
My Question now is, why didn't you just create UV shells for every Smoothing Group instead of creating more geometry around the edges, to get an result that isn't as clean as simply cutting the shells? Just because you didn't want to arrange too many uv shells?
On the left side you can see how i would do it. I think that also the way that Bek tried to say, he would do it.
(...)
My Question now is, why didn't you just create UV shells for every Smoothing Group instead of creating more geometry around the edges, to get an result that isn't as clean as simply cutting the shells? Just because you didn't want to arrange too many uv shells?
It's always going to be a trade-off between several things. If you have more UV islands you have more seams, which is less efficient (more splits*) and harder to texture (more seams), but potentially packs better. You can add extra loops (as seen here with the suppressor example) and have less islands, but at the cost of more distortion amongst those islands. When you're making your lowpoly and UV's you have to keep all these factors in mind, and decide what you want to minimise, and at what cost. Doing test bakes is the best way to evaluate competing methods in any given case.
I also think that you're worried too much about results being "clean" (whatever that means). How the normal map looks is mostly irrelevant - how does it perform in the target engine? Does it hold up or is the shading too extreme? The less geo you have to spend to control shading, then the more "use" you're getting out of your normal map (remembering that the main job of the normal map is to control shading while saving geometry).
*Because in this example less islands have been used, the loops really don't add that much extra geometry at all. If you're not clear on how seams and vertex normals (hard or soft/averaged edges) contributes to the overall geometry cost, consult the second link Mr. Bolton posted back on page 3. Otherwise maybe take this to a new thread so we're not clogging it up with discussion.
More UV shells = less actual texture resolution on the model (because of padding). As UV splits or hard edges basically count as an edge loop because in a game engine, verts along a UV seam have to be duplicated, thus increasing the tri count any way. The ideal solution would be balancing adding extra edge looks with custom normals to help control the smoothing as well.
This video explains the different approaches pretty well
Notice that the gradients on the normal map are only on certain parts of the model, that can help reduce artifacts and allow you to stamp decals onto the flat areas easily.
Good stuff Ben, the only concern in my mind right now is that the damage on the upper receiver looks way too blurry, as if it were a 1K texture or something. Really like it so far though, the modeling is excellent and the bakes are hardly distinguishable from the highpoly from a distance.
May I ask you to share some of your damage photos, or perhaps where are you getting them from? Looking to expand my photo gallery in some way or another.
More UV shells = less actual texture resolution on the model (because of padding). As UV splits or hard edges basically count as an edge loop because in a game engine, verts along a UV seam have to be duplicated, thus increasing the tri count any way. The ideal solution would be balancing adding extra edge looks with custom normals to help control the smoothing as well.
This video explains the different approaches pretty well
Replies
I really like the final look of your low baked mesh. I was wondering: do you re-model a low by boxes over your high, to make it fit for the bake? Or do you just reduce the mesh and remove all those edges straight from your high and detailled mesh?
I make detailed blockout meshes that serve as the basis for both the highpoly and lowpoly. The idea is to get primary and secondary forms in the blockout so that the HP only involves preparing subdivision and adding tertiary details and the LP only involves optimizing and unwrapping.
I always rush on having a good looking high mesh, and then reduce it, but it does not always perfectly match, so I have to tweak my normal etc.
I'll give that way a try
Thank you
It looks like that areas is flat in the mesh (maybe I am wrong), is there any reason to do it this way instead instead of just removing those loops?
First texture whack
I might also want to desaturate the brass on the barrel/gas tube a bit more.
No, they're great tools but I'm not using them. My textures are still Photoshop fill layers, masks, patterns, and adjustment layers. I've been using PS for a long long time and I like to think I'm fast and efficient with it. I reuse materials and processes between textures, etc. I've used Substance a bit and it's nice. Mostly it's a better interface for a lot of things that are possible - but clunkier - in PS. Unfortunately I've found it to be crash-prone and hardware-intensive. When that changes or I get a new computer with a better GPU I'll move over to using it more. They're marketing it aggressively and the industry is widely adopting it.
Most advice I could give is derived from things I've read over the years. If you want to get clean great bakes every time, it's important to understand everything from these sources:
Some distilled advice points I can give are:
About the Marmoset setup: it's just one point light, behind and above the camera, and inside the camera's transform hierarchy so it's always shining directly on the model from any point of view.
There you can take a look at it inside marmoset viewer,
On the bottom left, you can see the Polycount.
Looking at the whole weapon, it looks a bit patchworky with all the different materials, is that on purpose?
I prefer to think of it like this: A hard edge requires a UV seam, but a UV seam does not require a hard edge.
That's the most succinct way I can phrase it anyway. A hard edge requires a seam because to do otherwise would cause artifacts. A UV seam can have a hard edge for free because the vertices are already split (and it's usually beneficial to take advantage of that free shading change), but you're not required to do so. Smooth surfaces like a cylinder's side don't need that hard edge, so you can leave it smooth. It might not end up making much of a difference — but it's not the case that one UV island must "always" have hard edges at its boundary. But yeah, it works out that way 99% of the time anyway.
Also I've got a question and a comment about the magazine: I did one of these types just recently and I decided to have the small ribs as part of the normal map; you've added them as geometry. What are your thoughts on the pros/cons? I decided against it because I thought the extra detail there would be unnecessary and because it would make the UV mapping + LODs somewhat more fiddly. As for the comment, the size of the 7.62x39 round makes it look like the magazine is only a single-stack, perhaps because having two smaller rounds wouldn't sufficiently block vision of the mag interior..?
Another thought, did you have any trouble fully utilising UV space with some of the smaller modular pieces (like the grip or magazine)?
Kaine123 said:Yeah it's a good point and I think many of these textures will benefit from a second round of polish.
I don't mind altering the exact lengths of different barrels and handguards if needed to make them compatible. I think ultimately Rec C and Stock C will be the only parts that "have" to go together because of their unique junction incompatible with the other receivers / stocks (maybe I just won't make them).
Sure. i like UV shells = hard edges because your results will always be correct anyway and it's easy to automate "unwrap-by-smoothing" or "set-smoothing-by-UV-shells" to ensure consistency
They'd probably look fine normal mapped, I just left them in because at the time it seemed like the physical depth would be nice to have. The whole side of the mag is one big relaxed UV shell so it didn't really affect that process, and LODs would just be collapsing loops. In general the mag LP isn't very well optimized, I might go back and reduce it. I probably made it too thin also, should probably be flared out to look more capable of holding 30rnd.
Not really, but since they're all textured at 2k the texel density for different parts is all over the place, like the grip is many times more dense than the receiver for example. For a real production many of the less essential parts like the grip or barrel could be downscaled so it kind of works out though.
Maybe Stock C could just have its attachment point changed to be adaptable with the other receivers (not marging with the dust cover), it would only have to be shortened slightly from the concept.
Earlier in this thread you posted this picture saying all the is red, is one uv shell.
So your shading without an normal map and without the geometry you added should look like the one on the right:
Now your UV Layout for that piece should also look like the one on the right in the following picture:
On the left side you can see how i would do it. I think that also the way that Bek tried to say, he would do it.
No if we bake both of this examples with the same highpoly, we should get the same results.
And we get nearly the same results. But when you take a look at the normal map, you can see that the left side, where every hard edge is a UV shell, is way cleaner.
My Question now is, why didn't you just create UV shells for every Smoothing Group instead of creating more geometry around the edges, to get an result that isn't as clean as simply cutting the shells? Just because you didn't want to arrange too many uv shells?
I also think that you're worried too much about results being "clean" (whatever that means). How the normal map looks is mostly irrelevant - how does it perform in the target engine? Does it hold up or is the shading too extreme? The less geo you have to spend to control shading, then the more "use" you're getting out of your normal map (remembering that the main job of the normal map is to control shading while saving geometry).
*Because in this example less islands have been used, the loops really don't add that much extra geometry at all. If you're not clear on how seams and vertex normals (hard or soft/averaged edges) contributes to the overall geometry cost, consult the second link Mr. Bolton posted back on page 3. Otherwise maybe take this to a new thread so we're not clogging it up with discussion.
This video explains the different approaches pretty well
Here's how I would approach that part.
Notice that the gradients on the normal map are only on certain parts of the model, that can help reduce artifacts and allow you to stamp decals onto the flat areas easily.