Cubic Map Creation: Shader, Script, Workflow

[jedimasta]

My lead and I have been struggling with this for a few days now. We've been tasked with creating a QTVR-like (note, NOT QTVR) 360 panorama for a client. The intended target is for interweb use via a flash tool that simply requires 6 images, I'm learning, commonly referred to as Renderman or Cross-cubemap images.

Two Mental Ray lens shaders that I've located both do the job wonderfully: Mentalray Lens Library (http://www.atomlabs.org/asdn/mentalray/mrsl/lens/lensl_index.htm) and Lightengine3D Aperture (http://www.lightengine3d.com/index2.html), but I have two 'damned if I do, damned if I don't' problems.

Both shaders are strictly 32 bit Maya compatible only and I don't have the knowledge, nor the access to source files in order to recompile either for 64-bit processing, which our primary render machine runs - 64bit Vista with Maya 2008 64bit. Despite attempts to install .net framework, etc.. etc.. I continually get errors upon render.

Secondly, we have a secondary 32 bit render machine with 32 bit Maya 2008 that has no issues with the plug-ins, but the meshes we're using are high-poly count files that cause Maya to simply run out of memory on Windows XP (32 bit), halting our process entirely once more.

I'm stumped. I really LOVE the idea of a one-touch cubemap render to get the source images for our pano's, but aside from manually placing camera rotations at 6 keyframes and making the attempt to stitch, tweak and blend the images so that seams aren't formed, I'm not sure where to turn.

Open to any and all help. Thanks!

[gster123]

Hmmm...

I think I get what your after, converting a 3d scene into a cube map?

Off the top of my head could you not use something like HDR shop to do it?

That converts spherical, panoramic and cube maps maps such as HDR images into cube maps (or spherical or panoramics) pertty much in the touch of a button? You just need an input image to do it.

What you might have to do is have your environment make a sphere, up the reflectivity to 1 and just render the sphere, nothing else but make sure that everything is lighting/reflecting the sphere but its primary visibility is set to off, then use HDR shop to convert it?

It might be worth getting HDR shop (its free) and some spherical or panoramic HDR (or LDR dosent really matter for a test) images (free also from the same site if I remember correctly) and then seeing if HDR shop does what you want, as I might have got it wrong with what your after.

Hope its of some help.

Cheers

Steve

[jedimasta]

Thanks gster, but I think you misunderstand my intended end product. The pano I'm trying to output is a full 360 degress, not a single band or distorted sphere reflection. To clarify, have a look at the atom labs user manual: http://www.atomlabs.org/asdn/mentalray/download/mrsl_LensLibrary_en.pdf .

In the cubemap tools, you'll see what I'm trying to output. 6 distinct images that represent front, right, left, up, down and back. If mapped onto a cube from the inside (which is essentially what the flash element I spoke of earlier does) it gives the effect of being inside an environment and the camera can be rotated in any direction seamlessly as if you were sitting in the environment and twisting your head in whatever direction.

Both shaders I mentioned produce this 'cross' image without having the need for a complicated camera array or requiring me to reposition the camera, which I wouldn't mind doing necessarily, but have been unsuccessful at matching angles. Like I mentioned though, both shaders are 32 bit only.

[gster123]

Have you had a look at HDR shop, just to see, as I think its pretty close to what you want. You just need the right input for it.

[gster123]

http://www.hdrshop.com/

Have a look and go to tutorials and then Panoramic Image Transformations

[jedimasta]

No, I get that HDRshop can resample the images to the format this other flash tool would use, but the point is that I can't get the maya image (mental ray) in the first place.

To my knowledge, there are no native shaders/camera settings to create the type of image I'd need to have HDRshop do the conversion. Does that make sense?

[gster123]

Yes, unfortunatly that seems to be the case. Or at leas I've never seen anything like that.

The best way round I could see would be to render off one image (like a sphere) and convert that, then you dont have to mess with seams etc etc.

Not quitethe single shot that your after but possibly an easier way round than making 6 images and doing it manually.

Up to you really.

[jedimasta]

which begs the question... how do I get the sphere image in the first place?

[gster123]

I said how to get it in one of my previous posts. In addition your best off rendering as a square, or for a panoramic you could use a fish eye lense, not too sure if they work with 64 bit though.


I had a thought about this, not too sure if it will work though, as I've not checked it out as its just a thought. You could try making a cube in the scene, giving this a blinn and setting the reflectivity to 1, then baking all the info into a file texture via mental ray in the hypershade. As a poly cube's UV's are laid out as a cube map it should have the info that you need, in the baked file, but I'm not too sure if you have limitations with that such as raytracing etc etc, worth a look as it might solve the problem.