Compositing into a Cube Map

avenali312

Hey all,

I have a weird situation that has arisen. I have a set of six photos that have been turned into a cube map. The cube map is perfectly seamless as it sits. I've been tasked with compositing in some 3D elements into the cube map. This wasn't a big deal until I realized the items being composited span the seams.

I've attempted compositing into each individual photo in the cube and seeing how they line up, but it is far from seamless (the composited item resembles train tracks, so you can imagine all the railroad ties shifting perspective creates the issue). Manual Photoshop editing post-render is proving tedious.

Can anyone clue me into a alternative workflow that may make this a bit easier?

Sorry, but I can't share any of the images or content.

Thanks,
Adam

Farfarer

You could set up the 3D elements inside your 3D application then render them out and composite them back onto the cubemap.

Just render 6 images with a 90 degree FOV camera (up, down, front, back, left, right).

You could probably bring the cubemap in as an environment to your 3D app and check that all your 3D elements line up with the cubemap by looking through the camera. Then you can just render out the new cubemap faces without compositing it back on.

GlowingPotato

There is a really old but gold tool to help with cubemaps. Its called skypaint http://www.skypaint.com/

Or, as Farfarer told, youc an use any DCC tool to do this. like this tuto for 3DS https://judegodin.wordpress.com/2011/12/06/how-the-hell-do-i-make-a-cubemap/

avenali312

Thanks for the replies so far.

I'll try to explain a bit better. Our company sent a photographer out to capture a cube map with his, what I hear, very expensive camera setup. Those photos have then been stitched together using software, and converted into a cube map that I now have.

I need to then composite some 3D elements into the six photos that seamlessly span the images.

GlowingPotato

With skypaint, you can open your cube map in Photoshop and paint anything over it. So you need to render your objects in the perspective you want, with the illumination you want and then paint over your cube map.

But, maybe would be easier to create the scene in your 3D application and then render all images as Farfarer told. Then recreate the cubemap with those renders.

If the objects you want to compose needs to interact with the cube map, like reflections, and illumination, then you need to render those objects with the cube map as an environment.

Eric Chadwick

I've done this kind of thing a lot, in 3ds Max. It's fairly painless once you get the workflow figured out. Anything in 3D can be rendered into (or on top of) a cubemap.

I put some notes here
http://wiki.polycount.com/wiki/Cube_map

RN

These new 3d objects need to be rendered into their own cubemap texture.

With your 3d software you can map a cube with the original photography cubemap texture so you have a reference while building the scene, but when you're rendering you will hide this reference cube so only the new 3d objects are rendered.

Then you can bring both the original cubemap and the new, 3d-object-only cubemap to GIMP or Photoshop and composite these textures there. Ideally you won't paint or edit the new cubemap at all because you need to preserve the camera distortion introduced when rendering.

avenali312

Thanks everyone! I'm going to make another attempt today and see what happens. I'll report in again later this afternoon with my progress.