It's good to do your test process's without any Supersampler options enabled and at a smaller size until you have ironed out as many of the kinks in your setup's grab distance (offset max/min height) as possible, as these can be done more speedily. I often rely on not using the cage so that max's way of grabbing information is based more on firing off in a number of directions based on the normals of each of the lowpoly faces which is more similar to what is done in SHtools or more recent versions of 3D Studio Max. Sometimes this or comparable setup's to this can cause the cage to be skewed and changing the size of it will only cause it to become more skewed. Specifically, if you are optimizing your hi poly and low poly elements by removing all the areas of the lowpoly that are mirrored on the UV's, like if there was a rectangle and the long sides were mirrored, you had deleted the additional side so you have an open face. Often it can also be the cage itself that does not lend itself to a proper process. It's also real easy to adjust the power of the surface bumps just by adjusting the opacity of this layer. You can just overlay this over you original normal map in Photoshop to do the surface detail. We do a lot of the surface detail bump maps in Photoshop, using the Nvidia normal mapping plugin, it just takes a grayscale image and converts it into normal map. Then you can just overlay them, or use hard light on top of your processed normal map and adjust the opacity of the layer in photoshop to get the desired effect. All of our detail bump-maps are done this way. I usually set the nvidia filter from 20-60 and get a nice deep normal map. If it is too noisy undo and use a Gaussian blur on it to soften it a bit then run the nvidia filter again. Then just run the nvidia normal map filter on the image. This will create a cleaner normal map than just running the nvidia filter on an 8 bit image.

Play with the settings to see what works best.Īfter I've got that to where I want it, I convert it to 16 bit. This neutralizes the image a bit, and evens out the tones out somewhat. The process used varies slightly, but for the most part, I dump down the saturation (in hue/saturation) to greyscale and run a highpass filter on it in photoshop. The filter is a plug-in for photoshop and can be downloaded from nvidias website. I created those textures from photosource using Nvidia's normal map filter. When asked how some of the normal maps in Embry Square were created, Epic lead artist Chris Perna answers:

Normal map plugin photoshop