360-degree cube camera setup (updated)

Started by Matt, March 04, 2016, 04:11:01 AM

Previous topic - Next topic

Matt

[This is a new post with corrections to the file downloads]

If your version of Terragen has a spherical camera option (Terragen 3 Professional, Terragen 3 Professional with Animation), you don't need this setup to submit to the Terragen VR Challenge. Simply enable "spherical" in the camera node and render one image.

If you have a version of Terragen 2 or Terragen 3 that can't render with a spherical camera, then use this cube camera setup.

This is a set of 6 cameras to render a 360-degree environment in cube format. Each camera has a 90-degree horizontal field of view and represents one face of an imaginary cube surrounding the viewer. The cameras are named:

Cube Face Forward
Cube Face Back
Cube Face Left
Cube Face Right
Cube Face Up
Cube Face Down

Instructions for use:

(This is a lot easier than it looks, I'm just describing all the steps as clearly as I can.)

1) Download the clip file named "Terragen2_VR_cube_face_cameras.tgc" for Terragen 2 or "Terragen3_VR_cube_face_cameras.tgc" for Terragen 3.

2) In Terragen, go to File -> Insert Clip File, find the file you downloaded and open it.

3) If Terragen asks you whether to automatically insert the clip file, choose "Add Clip".

After a second or two, in the Node Network View you should see new node group containing 6 camera nodes. If you like to work in the Node Network, move this group somewhere close to your render node or other cameras nodes in your network.

If you click on the "Cameras" layout button at the top of the UI, 6 new cameras should also appear in Cameras list at the top left of the UI.

4) The cube faces cameras have the correct rotation and field-of-view settings to render a 360-degree view of your scene, but they don't know the POSITION of your viewpoint. Copy and paste the position coordinates of your original camera into the positions of each of the cube face cameras. You can do this with the clipboard button next to the camera's position parameter.

5) Change your render settings to render a square image. In your render node, set the image width and image height to equal values. For the Terragen VR Challenge, you can submit renders at 500 x 500 pixels or more.

6) Assign one of the cube face cameras to the Camera input. Now you're ready to render an image! To render a complete 360-degree view you will need to render 6 images like this.


Or, just watch the video  8)

https://youtu.be/JQt_FD_Ps28


Note about multiple cameras:

If you use the 3D Preview to view different cameras, take care not to use the navigation tools or navigation inputs when you're looking through camera that is not the current render camera. If you move the view, any changes will affect the camera directly, even without pressing the "Copy this view to current camera" button.

I prefer to do most of my scene building with a normal perspective camera, then connect the render node to the 360-degree cameras only when I'm ready to render the full 360. That way I don't accidentally change the cameras.

Matt
Just because milk is white doesn't mean that clouds are made of milk.


Matt

Just because milk is white doesn't mean that clouds are made of milk.

Clay

IS this only for the pro version and not the free trial version?

Matt

The cube setup works in the free edition. If you have the Professional edition you can use a spherical camera instead, which needs only 1 render instead of 6.

Matt
Just because milk is white doesn't mean that clouds are made of milk.

Dune

Speaking of which, is there a difference in render time between the 2 options, for the same result (detail-wise)? In the spherical projection the top and bottom ends are extremely stretched; does that have an influence on the sharpness/detail in those areas, in relation to the cube renders?

Clay

#6
Quote from: Dune on March 09, 2016, 02:18:09 AM
Speaking of which, is there a difference in render time between the 2 options, for the same result (detail-wise)? In the spherical projection the top and bottom ends are extremely stretched; does that have an influence on the sharpness/detail in those areas, in relation to the cube renders?
I am concerned as well on that actual issue, I've done a Helluva lot of QTVR renders from Bryce back in the day, worked with the guy from Apple computer that actually made quicktime and the QTVR to be exact, so basically your cam has to be "world centered" in your scene, and yeah do the 6 images to "cube render", but yes you should "in theory" use the same settings as usual, I would think for all the light and defer atmo etc. I'll see if I can dig up some old Bryce ones for an example, again I might be off my rail here cuz I'm still learnig this app:-) Neeevermind those are on an old server I can't get to at this time LOL, but there use to be a Mac App call Make cubic, I wonder if thats still around, there's plenty it seems for windows users but not much for Mac peeps, and it was a free app///hmm interesting.

Matt

#7
Quote from: Dune on March 09, 2016, 02:18:09 AM
Speaking of which, is there a difference in render time between the 2 options, for the same result (detail-wise)? In the spherical projection the top and bottom ends are extremely stretched; does that have an influence on the sharpness/detail in those areas, in relation to the cube renders?

The cube map is more efficient. In the cube map, some directions have a higher pixel density than others and the pixels are slightly distorted, but these variations are acceptable.

In the spherical (equirectangular) projection, as you know, the pixels get more and more squashed as you get closer to the poles at the top and bottom. The vertical resolution is constant but the horizontal resolution increases. Near the poles it has more horizontal resolution than vertical resolution (and more than at the horizon), so in that sense it's wasteful.

An aspect ratio of 2:1 is designed to create square pixels near the horizon (and for this purpose the spherical map is better than the cube map, but only near the horizon). At a resolution of 2000x1000 the pixels at the horizon have the same resolution (on average) as a cube map with 500x500 on each face. This cube map would have 1.5 million pixels, while the spherical map would have 2 million pixels. That's 33% more pixels.

In terms of render times, you might expect that a spherical at 2000x1000 would take 33% longer to render than a cube at 6x500x500. But I don't know if that's true in practice.

Finally, it's true to say that the spherical image does have more information. The 33% more pixels capture more horizontal resolution, after all. It's not how you would choose to use those pixels in an ideal world, though.

Matt
Just because milk is white doesn't mean that clouds are made of milk.

Clay

So, using the free beta test version what is the "ideal" image size to make the 6 image cubical image? square fisheye etc etc

Matt

Hi Clay,

In the video ( https://youtu.be/JQt_FD_Ps28 ) you can see that I used the camera setup in the .tgc file which you can download from the first post. Use these cameras, and don't change any of the settings except for the position. To submit to the contest you can render each cube face at 500x500. You can go higher, but they must be square images.

Matt
Just because milk is white doesn't mean that clouds are made of milk.

Clay

Quote from: Matt on March 09, 2016, 08:52:46 PM
Hi Clay,

In the video ( https://youtu.be/JQt_FD_Ps28 ) you can see that I used the camera setup in the .tgc file which you can download from the first post. Use these cameras, and don't change any of the settings except for the position. To submit to the contest you can render each cube face at 500x500. You can go higher, but they must be square images.

Matt
I shall give it a whirl :-)

Clay

Hey can one of you guys help me a bit? I rendered my 6 images but I cannot find a decent Mac OSX QTVR processor app that I'm comfy with, make cubic for mac is only for PPC and not for intel based machines now sigh..


HecubaAssassin

Stereogram in Paid Edition?  Nope, sorry nope.  I have the Pro without video from launch of v3, and it is not in the camera settings that I can see.  I've downloaded 4 or 5 photostitchers and didn't get them to work.  Has anyone else had luck with finding software for stitching the 6 sided cube?