Normal Mapping Issues

[Deleted User]

Can someone please tell me why I'm having issues with this normal map? I've attached a small test model where I got perfect results, using the same method that produced the issues on my main project. In the past, I used a lot to BnW bump maps, so I never learned much about normals.

It's not even a complex model.

Software:

• Max 2014
  
• Handplane Baker
  
• Marmoset Toolbag

Things tried

• Single smoothing group
• Multiple smoothing groups
• FBX, OBJ, triangulated, and quad outputs

Obscura

You forgot to mention what's the issue.

[Deleted User]

Sorry. Loot at the gif above. None of the model's features are recognizable in the map because everything jumbled? How come this map turned out fine and the one above didn't?

[Deleted User]

Obscura

You mean the "line" is missing on the edges, but they are there in the corners? 
If its that then I'd say its probably fine... You have support loops on your lowpoly so the most of the normal map gradients will get flattened out. Have you tried to apply the normal map to the mesh and see if it actually works?

[Deleted User]

Obscura

"None of the model's features are recognizable in the map because everything jumbled"

As you can see on your preview when its applies, it does its job... They are now really showing in your normalmap, because every of the features are also there in the geometry...If you'd remove those details from your lp mesh, they would "show up" on the normalmap. On your another example, you bake the details to a planar surface...While on your example that has the "issue", you modeled everything into the lp so you bake only edge smoothness.

So with this low poly mesh, its normal to get that look on your normal map.

huffer

Also your low poly mesh could be heavily optimized, you don't need another row of loops on every inset if you want to bake a normal map. With that much geometry, you don't even need to bake a normal map, keep everything one smoothing group and with support loops the edges will hold their bevel.

 You can remove all support loops on insets and edges (keep just the edges that maintain the shape) and add that geometry in roundness - your rounded corners have only two sides (is that even the low poly in the gif or the unsubdivided high poly?). That's the cause of some artifacts you see in the preview - the low poly should match the high poly as close as possible, especially with round geometry.

I suggest reading the polycount wiki entry on normal maps.

ActionDawg

A normal map is storing the difference between your high and low poly models. Thus if they're very similar the changes you see will be subtle.

[Deleted User]

huffer said:

...is that even the low poly in the gif or the unsubdivided high poly?

It's the undivided hp model that also served as a test lp. I jumped the gun and started screwing with maps.

AlecMoody

A few things.

First, try to think of normals baking as two seperate processes. Projections, and shading behavior. It's a lot easier to see what your projections are doing with an object space bake so first bake an object space map. Load that into the model in a 3d program with no shading/lighting.
The object space map makes it really easy to see what is going on. The tangent space maps is kind of a jumbled mess. Ray misses in handplane baker will be mid grey. We will add an option for ray miss debug color in the



Second, try increasing the backray offset scale in the settings here:
http://i.imgur.com/qtEAk4R.png

A default of 1.0 sets in the inner ray length to match the outer ray length. For very tight cages, this is often too low. I will probably switch the default to 5 on future builds and I plan to make a video on setting up and controlling projections in handplane.

[Deleted User]

So, the artifacts are gone...along with everything else. Nothing is baking and i just get this...