Looks fukin awesome. Left pic looks crazy believable.
Is it possible to get a modo vs. in engine shot? presuming the last shots were in marmoset. I'd be majorly interested in seeing a comparison if thats cool
Really great details, but is there really 12k tris on the ut3 weapons, i have been looking at the vehicles quite a lot, and most of them are around 10 - 12k would they really have more tris on the weapons than on the veichles? not that it's a problem or anything, just got me a bit interested.
[ QUOTE ]
Really great details, but is there really 12k tris on the ut3 weapons, i have been looking at the vehicles quite a lot, and most of them are around 10 - 12k would they really have more tris on the weapons than on the veichles? not that it's a problem or anything, just got me a bit interested.
[/ QUOTE ]
Dfacto was telling me they were around that range. Really its sort of interesting if you think about it. In a FPS game the weapon is the only(or atleast primary) means for interaction the player has, that is litterally who he is and how he communicates with the world. Not only that but it is the only thing that is up close in your view 100% of the time, making it the most important *character* in most games. Atleast thats how i look at it, its a really good way to connect with the player and create a sense of immersion, without good weapon models and animations the entire experience can feel flat.
Most of the weapons i do actually have higher specs than the characters in our game, simply because most characters you wont be staring at up close, and will be moving around motion blured at most times, and you just wouldnt notice the increase in detail. But on the other hand, you see the weapon models up in close view constantly, and a lot of the time standing idle, so it has to hold up.
You can look at games like killzone, now really if you strip out the kickass weapon models and animations its a shit boring really standard shooter. But with the extra time spent on those assets it actually makes it more fun.
iirc the epic guys were saying the exact same thing in the making of video available in the CE edition of UT3, they treat weapons as carefully as characters and try to give them as much detail as possible because they are the real heroes/avatars of the game. Or maybe was it in the CGTALK interview...don't remember exactly
Ok, heres the flats if anyone wants to see, scaled to 512 from 2048. Could have mirrored stuff if i wanted to use tangent maps, and still might re-organized and convert em but oh well, works for now.
Why Would you use object-space over tangent? Won't using Object space keep you from animating the clip (unless you baked that separately and kept it as a separate object)? Same with the charging handle, etc.
I haven't used object-space maps much and while I understand they're useful for static meshes because they're somewhat faster, It seems like a weapon like this would benefit from using tangent space.
No as long as your engine supports it there isnt really any animation limitations with rigid parts(soft deforming objects *may* be ok as well, we havent really had the need to test this much) the engine has to keep track of how the rig is transforming the verts, so it can keep track of how the lighting needs to be transformed on the mesh.... Someone else can explain this better, but it works
Is tracking the vert transforms and adjusting lighting still more efficient than the rendering hit for tangent maps? Or is it less a matter of efficiency and more a matter of relieving GPU load by increasing CPU load?
Slum: quality and simplicity of workflow, its a lot easier to get really good results using object space, you can add as much geo to your cage mesh to make sure things render out straight without worrying about screwing the smoothing with the end result. The main thing is just getting it to render out a really nice, clean normals map for this hard edged stuff. Its really a pain to spend days modeling out a nice smooth highres mesh, only to have to chop up the lowres mesh with smoothing groups and get nasty seams and such. I've played with the object space to tangent converter in xnormal which is really cool but still not quite the same quality, atleast without redoing the uvs and adding in gaps at every smooth group so that the edge padding will actually fix the aliasing you get at smooth group borders....
I really dont think its much faster or slower than tangent, i doubt its a noticable difference either way. Also since you dont have tons of gradiation over your maps compression actually works quite a bit better with object space it seems.
Ahh been meaning to ask you how the texturing was coming along, guess you saved me the trouble
Looks good to me, and the dials really look hot on the scoped view.
Only thing that I find odd is the lack of color/material variation for the idents on the upper sides of the scope.
I finally understand why you're using object space, no need to slice up your uv's like a crazy dog, though I'd still be using tangents with my masochistic technique until kingdom come
And yeah, object space is awesome sometimes, especially for shit like this. If you used tangent space, there would likely be a good deal of shading issues. This way, it's 100% lighting from the high res, no smoothing group worries. Clean and sexy.
The big downside to object space though, is that you can't add detail in post, like you can with tangent space maps. So you had better make sure your high poly has all of the detail you want in it.
Yeah we have a tool that lets you combine a tangent space normal on top of an object space one, you generally need to export a mesh with split up smoothing, or else the results get sort of "bent" around the mesh's normals tho so its a little buggy...
Xnormal has a tang->obj converter, i
think you could probably do this as a major hack around it:
Render out object space map
set up smoothing on a copy of your mesh
convert to tangent space
add normals detail
convert back to object
compostive image back over the orig object space map, painting out problem areas (like edges from smoothing, etc)
but that would be a bit of a pain in the ass
I actually did add that text detail in the normals as post work in PS, it can still work at times, usually only on the sides of objects, and you have to manually get the oreintation right by inverting channels and such, so useless if you wanted to add some bump noise to over a curver or something.... Well, imo you dont really need that stuff if you've got a good diffuse and spec map, and in most models you see its overdone anyway.
So yeah, main disadvantages are:
You cant mirror anything
Adding extra bump detail is a bit of a pain
Is that what that panel is for? Adding a bumpmap for the lowpoly's uvs? Or is it for the highpolys? I've never really understood what that was for, infact i thought it was a feature you abandoned a while ago.
In any case it would be great if there was an option somewhere to combine a bumpmap/normals into an object space map, either while rendering or afterwords!
I actually ignored this thread for a while, until I noticed it was consistently at the top and had grown considerably. Nice work eq! I love the in-engine screens, the specular and colours really work fantastically.
Replies
Is it possible to get a modo vs. in engine shot? presuming the last shots were in marmoset. I'd be majorly interested in seeing a comparison if thats cool
needs more rust
nice going!
Really great details, but is there really 12k tris on the ut3 weapons, i have been looking at the vehicles quite a lot, and most of them are around 10 - 12k would they really have more tris on the weapons than on the veichles? not that it's a problem or anything, just got me a bit interested.
[/ QUOTE ]
Dfacto was telling me they were around that range. Really its sort of interesting if you think about it. In a FPS game the weapon is the only(or atleast primary) means for interaction the player has, that is litterally who he is and how he communicates with the world. Not only that but it is the only thing that is up close in your view 100% of the time, making it the most important *character* in most games. Atleast thats how i look at it, its a really good way to connect with the player and create a sense of immersion, without good weapon models and animations the entire experience can feel flat.
Most of the weapons i do actually have higher specs than the characters in our game, simply because most characters you wont be staring at up close, and will be moving around motion blured at most times, and you just wouldnt notice the increase in detail. But on the other hand, you see the weapon models up in close view constantly, and a lot of the time standing idle, so it has to hold up.
You can look at games like killzone, now really if you strip out the kickass weapon models and animations its a shit boring really standard shooter. But with the extra time spent on those assets it actually makes it more fun.
LOL Graphics = gameplay!
Almost feel like you could reach out and pick it up. Sexy work EQ.
Ok, heres the flats if anyone wants to see, scaled to 512 from 2048. Could have mirrored stuff if i wanted to use tangent maps, and still might re-organized and convert em but oh well, works for now.
I haven't used object-space maps much and while I understand they're useful for static meshes because they're somewhat faster, It seems like a weapon like this would benefit from using tangent space.
I really dont think its much faster or slower than tangent, i doubt its a noticable difference either way. Also since you dont have tons of gradiation over your maps compression actually works quite a bit better with object space it seems.
Looks good to me, and the dials really look hot on the scoped view.
Only thing that I find odd is the lack of color/material variation for the idents on the upper sides of the scope.
I finally understand why you're using object space, no need to slice up your uv's like a crazy dog, though I'd still be using tangents with my masochistic technique until kingdom come
Always a pleasure seeying your work
And yeah, object space is awesome sometimes, especially for shit like this. If you used tangent space, there would likely be a good deal of shading issues. This way, it's 100% lighting from the high res, no smoothing group worries. Clean and sexy.
Xnormal has a tang->obj converter, i
think you could probably do this as a major hack around it:
Render out object space map
set up smoothing on a copy of your mesh
convert to tangent space
add normals detail
convert back to object
compostive image back over the orig object space map, painting out problem areas (like edges from smoothing, etc)
but that would be a bit of a pain in the ass
I actually did add that text detail in the normals as post work in PS, it can still work at times, usually only on the sides of objects, and you have to manually get the oreintation right by inverting channels and such, so useless if you wanted to add some bump noise to over a curver or something.... Well, imo you dont really need that stuff if you've got a good diffuse and spec map, and in most models you see its overdone anyway.
So yeah, main disadvantages are:
You cant mirror anything
Adding extra bump detail is a bit of a pain
just release your engine so we can play with it
In any case it would be great if there was an option somewhere to combine a bumpmap/normals into an object space map, either while rendering or afterwords!
and happy bday