Hi Everyone,
Our texturing process is built upon baking in lighting information and data into a flat diffuse.
I'm running into an issue, where normal maps when doing a diffuse render to texture are providing incorrect surface information.
It may be easiest for me to illustrate via some simple imagery.
Here is a simple viewport render with a simple reflective material and HDRI background:
www.leatherneck-sim.com/RTT_Bug_01.jpgHere is the exact same setup, rendered to texture. Note the seams on the front.
www.leatherneck-sim.com/RTT_Bug_02.jpgThis is only the most obvious area in which this happens. It's more subtle in others, based on what influence the normal map should have to the model and mesh itself.
Here is a image for reference of the complete asset:
www.leatherneck-sim.com/HUD.jpgHas anyone run into something similar? Why does the normal map not work appropriately in the RTT setup?
This is with VRAY 3.2.
Many thanks in advance for any help,
Nick
Replies
I would look into Vray's normal map support, what tangent space it expects.
It could also be as simple as inverting the green and/or red channels.
I've tried flipping channels, and using both "Normal Bump" and "VRAYNormal Bump".
However, seeing as the normal map renders perfectly in a normal camera view, it seems that this may be something related to RTT and the projection it uses?
You're using the normal map in a V-Ray material, then baking a Complete map? Have you tried a Standard material with Normal Bump map and the Scanline renderer then RTT?
Vray in 2012, 2014 has this issue.
Mental Ray in 2017 has this issue.
I can't quite believe that Normal Maps cannot work properly in a simple RTT setup. Has anyone had similar issues?
Thanks for helping Eric!
The curious thing is that Mental Ray exhibits the same issue in 3Ds Max 2017.
If I unify the normals on the part in question, I get somewhat of the same issue in the viewport render. See image below.
www.leatherneck-sim.com/Issue_01.jpg
Here is a more illustrative view of the issue in the RTT window:
www.leatherneck-sim.com/issue_02.jpg
This leads me to believe it may be something smoothing group/normal related?
Perhaps there is a difference in how the surface normal is used in RTT? It must be common across Mental Ray, VRAY and Corona.
I believe Scanline uses a Camera based projection (You can get directional reflection and highlight bakes by moving your camera and viewport around), and perhaps that's why it is unaffected.
I've tried splitting the polygons away from eachother entirely, that did not help. Neither did unifying them together.
Any help is greatly appreciated,
Nick
I've tried with no maps except just the normal as well-- the same issue is present. (The first post actually has images with just the normal applied)
I'm starting to assume it's something due to how the renderer traces rays from surfaces and how that is different in RTT mode (flat out from surface vertex/normal vs camera direction based as in the viewport)
The environments are also set to their appropriate maps (VrayHDRI for Vray, etc.)
All maps are properly gamma corrected too.
I'd be very inclined to think it's a map or material issue if it didn't render correctly in the viewport/camera, however. It's literally 100% perfect, it's just when doing a RTT that the normal is incorrectly used.
Many thanks for the prodding though; you're prompting me to double check everything.
http://forums.chaosgroup.com/showthread.php?89164-Why-does-Vray-bake-normal-map-in-sRGB