I'm setting up an Octane camera rig to make stereocubemaps manually. Like 12 seperate renders. I'm having trouble maintaining seam alignment though. I have 2 cameras, 2.5" apart, and then they rotate 90 degrees and render again. This happens a few times to get all 12 shots. (yes I Know there's a stereocubemap option that will do it all in one shot). I have a little max script to go through all the cameras and render with different file names. Cam1, Cam2, etc.
So regarding the seams, where should my cameras be? How can I get them to connect seamlessly once they've rendered?
12 virtual camera rig
Hi DartFrog,
not a Max user sorry, but in Standalone is possible to choose the cubemap panoramic camera and render only one face at a time. If you don't find the option in max, you can export in orbx to standalone and setup the render there to avoid the seams
ciao beppe
not a Max user sorry, but in Standalone is possible to choose the cubemap panoramic camera and render only one face at a time. If you don't find the option in max, you can export in orbx to standalone and setup the render there to avoid the seams
ciao beppe
Well the reason I need 12 separate shots is because I'm also doing a pass using Mental Ray. Mental Ray won't recognize the Octane Stereo cubemap camera so I have to it with regular octane cameras, which it will recognize and use. So the problem is the stitching of the images. There are seams of mis-alignment that I'm having trouble getting rid of. I don't know how the octane cubemap camera works exactly because their stuff is all seamless, somehow.
Have you tried to go in the opposite way?
In theory, if you render in mental ray with classic equirectangular panoramic camera, then you can load it as an environment in an empty scene and render it in octane cube map even with alpha in one go
ciao beppe
In theory, if you render in mental ray with classic equirectangular panoramic camera, then you can load it as an environment in an empty scene and render it in octane cube map even with alpha in one go
ciao beppe
That would work great, but I need it in VR. So the Octane camera would be rendering a stereo image of a non-stereo background 
I can get everything to work if I use single cameras (12 of them) for both octane and mental ray, but then there are seams since I don't know how to stitch them the way Octane does.
I think I'm going to just not do it I really wanted it to work and get a pipeline going, but I'm going to re-think what I'm doing.
Thanks for your help.
I can get everything to work if I use single cameras (12 of them) for both octane and mental ray, but then there are seams since I don't know how to stitch them the way Octane does.
I think I'm going to just not do it
I really wanted it to work and get a pipeline going, but I'm going to re-think what I'm doing. Thanks for your help.
I think your idea of omni-directional stereo might be slightly off. To entirely avoid seams you need to basically capture one ray per pixel. It's not a discrete amount of perspective cameras that are then stitched together in a second step. That will never work unless you have a really large number of cameras and only use the central columns of pixels for stitching.DartFrog wrote:there are seams since I don't know how to stitch them the way Octane does.
Even people that do it with physical cameras are rotating the pair on a plate and capture video, or have a lot of them around the plate to try to minimize seams.
Maybe have a look at this: Synthetic Stereoscopic Panoramic Images (Paul Bourke, 2006). Especially chapt. 2.
Note that he doesn't mention the fact that even with this technique the stereo would be wrong at the poles. You need to merge the poles of the stereo pair, there is currently no way around it. See the "eye distance falloff" parameter in Octane. This is a continuous function of latitude so that's another thing you wouldn't be able to do with just a few discrete perspective images.