Someone asked how to render for 3D TV on G+ and I spent quite a while writing out a long comment about how it can be done and what you should pay attention to and so on and so forth.
Then I got a bit cocky and thought I’d try and do an ASCII art stereo pair.
Then I hit ctrl+shift+left arrow and my browser went back and lost my comment.
I whimpered at that point.

So lets have another go.

3D footage isn’t super mysterious really. The information needed to display a 3D image is simply a view from the left eye and a view from the right.
By presenting each of these images to their respective eyes, via shutter glasses for example, or the good old cross-eyed viewing method, the effect is that your brain perceives depth in the image.

So in order for you to render your fantastic new animation of a jelly having an anvil dropped on it, IN FULL 3D!!!!oneone!!11, you need only have two cameras.

Quick(ish) test. Load up your favourite 3D package (It’s Blender isn’t it? Isn’t it? I bet it is.) and render out a scene. For this experiment it makes sense to have a square or portrait orientation.
Save the image.
Go back to the scene and shove the camera over to the left or right (with respect to its orientation) by a few centimeters (keeping the focus of the camera the same if you can, but for anything but close foreground work, this probably won’t matter too much) and render it out again.
Save that image and open up your favourite graphics package (It’s GIMP isn’t it? Isn’t it? I bet it is.) and position each of those images side by side. Zoom out if necessary (it might be at first but with practice you can do pretty big stereo pairs) and go cross-eyed such that your left eye is looking at the right image and your right eye is looking at the left image.

Where the two ghostly images come together in the middle, you should get the depth effect. If the effect seems to be reversed i.e. things that were closer to the camera look further away than they should be) swap the order of the images round and try again.

For extra credit, do it on a piece of paper. Get a pencil and draw two boxes of as close to the same dimensions as you can, side by side.
In the boxes draw the same object in each, but position one of them slightly off to the left or right with reference to its box.
Go cross eyed such that your left eye is looking at the right hand box and your right eye is looking at the left hand box and prepare to have your mind blown (until you realise that it’s pretty simple optical mechanics really, and then you’ll calm down a bit, and then you might wonder how your brain does all this stuff on the fly and your mind will once again be blown, only this time it’ll be due to contemplating the power of your mind… which is blown… meta or what?).

It was at this point that I tried to do the ASCII art demonstration in a G+ comment and I buggered it all up.
I shan’t be trying that this time.
Maybe once it’s saved and published.
I might come back with some Blender made stereo pairs also.
Edit: Here we go! An example! And you can quite plainly see that it is in actual fact, a sailboat.


So in theory, all you need to do is render your scene/animation from two viewpoints to represent the two eyes in the head of the person doing the looking at, and present them to the two eyes of the person viewing the scene/animation.

In practice, exactly how you achieve that last step is down to what you want to use to view the image/animation. I have no clue whatsoever how your particular 3D TV expects its data to be presented, but the answer will be out there somewhere.
I know that NVidia’s 3D vision system is pretty flexible about whether you put the images side by side, over/under, or interlace them in some way, but personally I prefer side by side, right eye first. That way I can test it by just going cross-eyed.

Some people are able to diverge their focus, so the left eye looks at the left image and the right eye looks at the right, but I find that very difficult to do on command. Much harder than the alternative. In fact, the only reliable way I’ve found to do that is to actually look at something much further away than the image then slide the image into view. This works a treat with those old magic eye books because you can look over the top of the book out the window then bring the book into view and relax into the right separation for the image.

Why they tended to be mostly¬†divergent type, I’ve no idea.

Incidentally, those magic eye pictures are random dot stereograms. They don’t use two images as such, they use vertical strips of a repeating but semi-random pattern that is then “shifted” left and right in the shape of the object they want to show in depth.

So there.

That’s how.


(edited to remove split infinitive)

Leave a Reply

Your email address will not be published. Required fields are marked *