Since there really arent a lot of resources on this subject i figured i'de start a thread, heres a lil thing i made for TK to show him how to do soft intersections with floating geometry. Its not always perfect but works pretty well and you can touch up the spots that dont look right in seconds in ps after you render normals.
[edit] Ill explain waht im doing here a little too, extruding/beveling/whatever extra edges out of the shape perpidicular to the plane of the surface which you are trying to add an indent/extrusion to so that you dont get a hard intersection. It basicly will just match up how it shades and fool the normals renderer into thinking its on the same plane so you dont get a hard edge.
[edit][edit] All floating geo will cause some problems when rendering ambient shadows of course, but that is generally easy to paint over in ps.
So feel free to contribute everybody. And i can post some more stuff if people have specific questions.
Replies
As for the secret "make high-res art" button its a max only feature malcolm
I'm sure anyone with high poly skills COULD make the model water tight, if they wanted to steal money from their employer by being as wasteful of time as they possibly could to satisfy their own artistic ego, but for those of us with the desire to be efficient, and to get more assets of our own into the game, I highly suggest figuring out how to fake detail without welding it into the high poly mesh.
To the normal map baker, it all looks the same, and the high poly never goes in the game, so there really is no downfall whatsoever to this creation method.
Cgtalk Subd thread
modeling hi poly weapons is painful, any tips?
Modeling water tight high poly mechanical models (5 times fast) is a lot like like going insane, it takes time, dedication, perseverience and ingrediant X.
If so then Toomas' idea might work. Seems like extra work to me, but I don't do that kind of rendering anymore. Though retleks seems dead-on about the perseverance, gotta be into that kind of noodling, just like any other discipline I guess.
why not just paint the normals manualy? you can go back and push it further, make sure it's sharp as a tack, no artifacts are possible..
just grab a sphere, put it on a grid in 3d studio max, generate the normal map from the grid, and color sample all the little facits of the sphere. you could paint that sword onto a grid, and it would look awesome in 2 minutes. i'd also like to point out that in a lot of cases some models don't have enough unique UV space to take advantage of floating geo. i find that doing little painted normals tricks is a godsend.
obviously this totally depends on your polycount, and what you're making... but personaly i think this whole high poly geo thing is out of control. i've seen people opening up zbrush to sculpt in little groves where 5 minutes and extremely good use of the nvidia photoshop filter would have created an overlay that would have made it impossible to tell the difference.
But I'd love to see how you would paint the sword normal map and still make that tip come out well. I can see how you could easily make a decent gradient for the shaft, but getting smooth transitions to the tip and to the base of the shaft are tough to paint well.
Also about floating geo bit, really one of the biggest adfantages is this. Say you have a slightly organic mechanical shape with lots of curves in it, your uv is never going to be perfectly undistored, so adding details after the fact will get you distortion when its displayed on the mesh. I remember someone having some odd solution for this but really when you have to counteract things like that you really are spending much much more time editing those details than it takes to just model them.
in my opinion, UV space does matter. I'm finding that when i work with super low texture sizes especially, and a lot of geo shares the same texture space, you can't get away with a lot of the straight up generate-from-high-poly technique because there isn't enough unique texture space to bend the normals across a large surface. there are cases like that, i deal with them allll the time, and often times i find that if I've very clever with my UV space, i can paint in little beleved edges here, little slopes in the normal map there, and still get away with using a lot of duplicate UV space. does this sound that crazy?
i also think that for certain objects, it is easier to hand paint the normal map. and where do you get off making the assumption that i think that this is a way to handle complex detail? did i make the statement that i think ken Scott should have hand painted the doom models?
I've seen some AWESOME stuff done with hand painting normal maps on SIMPLE geometry, and if it's done properly it can take a fraction of the time. apparently you made the jump to assume that i was referring to all normal map creation in history & the future..
c'mon guys.. whatever
(post for comic relief only)
or when you want to add nuts and bolts.
you all suck, i win!
both techs has its pros and cons... The discussion is pointless. For some stuff modelling is better and for some s-filter is the way to go. Sure we can enlighten people in what situation one method is to prefer rather than the other but now point goin all crazy about it.
you all suck, i win!
[/ QUOTE ]
The obvious truth is stated, duh guys!
btw wazzup EQ.