Could use some critique and pointers on my normal map
polycounter lvl 14
Hey guys,
This fire hydrant model is my first serious attempt at normal mapping and I could use some critique and pointers. I've been through the normal maps threads and have learned a ton, but something's still not right.
On top of the pointers/critiques, I do have a couple of questions ready:
1) Was it okay for me to normal map each part of the hydrant separately? The high poly sculpts were easier to work with that way, but now that I'm looking at it, could the distinct shadow drop off (say, from the front hub to the body) be caused by this approach? It seems like an obvious question that can be answered simply be testing it, but if this is the case, how do you guys handle extremely dense meshes? Surely there are cases were separate bakes are required for the same object?
2) UDK really flattens my normals out. They show in Unity much better. What gives? I made sure to invert the green channel for the UDK shot.
3) The waviness is more apparent in the Unity shot. Let me guess, add more geo to low poly? Besides letting the unwrap for the cylinder become a banana shape, I don't see what else I can do on the UV end of things. Plus, I remember seeing Earthquake specifically state that banana shaped UV's are "retarded." I've aligned the UV's horizontally and let the vertical lines distort a bit. I'm pretty sure the answer is going to be add more geo... but man, I really thought I 16 sides would have been enough. I think the lesson learned here may be, the bigger a cylinder on the screen, the more sides it will need to read correctly.
This fire hydrant model is my first serious attempt at normal mapping and I could use some critique and pointers. I've been through the normal maps threads and have learned a ton, but something's still not right.
On top of the pointers/critiques, I do have a couple of questions ready:
1) Was it okay for me to normal map each part of the hydrant separately? The high poly sculpts were easier to work with that way, but now that I'm looking at it, could the distinct shadow drop off (say, from the front hub to the body) be caused by this approach? It seems like an obvious question that can be answered simply be testing it, but if this is the case, how do you guys handle extremely dense meshes? Surely there are cases were separate bakes are required for the same object?
2) UDK really flattens my normals out. They show in Unity much better. What gives? I made sure to invert the green channel for the UDK shot.
3) The waviness is more apparent in the Unity shot. Let me guess, add more geo to low poly? Besides letting the unwrap for the cylinder become a banana shape, I don't see what else I can do on the UV end of things. Plus, I remember seeing Earthquake specifically state that banana shaped UV's are "retarded." I've aligned the UV's horizontally and let the vertical lines distort a bit. I'm pretty sure the answer is going to be add more geo... but man, I really thought I 16 sides would have been enough. I think the lesson learned here may be, the bigger a cylinder on the screen, the more sides it will need to read correctly.
Replies
will look much, much better
To get rid of waviness you need to use a bake mesh that is different from your low poly. Make your bake mesh denser and expand the mesh around your high poly. To do this just select all your vertices, get the move tool and move them out by the vertex normals. In Max I think you can use the push modifier to do this. This will allow the rays to capture correctly. A way to manually fix the waviness is to select a small portion of it in photoshop and just stretch it out in the direction of the waviness. This won't work on wavy parts that are circular. Would be super awesome if photoshop had a radial transform or something.
Would anyone be so kind as to do a paint over on my wires here? I'm out of ideas on how to reroute edges so it doesn't cause normal issues. I think I've redone the geo roughly six times now, and I just keep getting the same results.
Any more protips?
I really like how the bevel is working out now... except the part where the shading is off.
For my high poly, I'm essentially just duplicating the low and adding control edges to bring it into Zbrush to crank the divisions up on, and sculpt on of course. Is this not sufficient? In xNormal, I have to expand the cage a little and save the meshes from there.
EDIT: Didn't realize how small the image was. Here's another: