Good stereo 3D - Questions

I am experimenting with stereoscopic rendering in ToonBoom Animate Pro and already made a camera rig following the tutorial in the resource section. After being very impressed by the first awesome results, I realized that for a real productive use there is a ton of stuff more to figure out:
1. The camera set up in the tutorial is looking inwards at the focal point = convergent. What do I do with far away backgrounds? How can I find out the point where the background (or foreground) parallax effect is getting ripped too far apart. I have seen some rigs in 3D programs that let you set up the closest, focus and background point (after which the background image is more or less flat). Is something like that possible in Animate Pro?
2. Also I have a big problem with distortion due to the convergence. Even the layers being exactly on the focal point have a huge parallax at the sides (even going up and down) and no parallax at the center (how it should be). It still kinda works… but is there a way to improve this?
3. Straightforward views are pretty easy, but as soon as the camera is looking in an angle (e.g. slightly looking down so that the horizon is not that low) I need to rotate every object so that they face the camera to not reveal their cardboard nature. Rotating every object by hand is tedious. Is there a way so that certain layers always turn towards the camera?
4. What about a parallel camera set up? Does it lead to everything being in front of the screen in any case?
5. What else is important for a good stereo 3D experience?
I would be very glad if some people could share some of their experiences about stereoscopic images and maybe help me to figure out some of the questions.

Thanks for the answers :slight_smile:
That link is really helpful… a great overview about everything you have to keep in mind for stereo 3D.

The ortholock module is very helpful! Great!
Yeah I guess I can find out the 3D comfort zone by trying. It’s just that… in theory you could calculate the maximum near and far parallax preciously. Ideally you even have to render different versions for the TV and the cinema… and I wish that could do everything in Animate Pro without having to change the actual distances of the objects… Guess instead I could change the cameras distance to create version with more and less maximum parallax… I have to try that.
The keystone effect with the values given by the tutorial in the resource section at zero parallax is huge though. The ortholock helps I assume since it eliminates the rotation… I will try to change the eye distance.
Rendering the scene with parallel cameras and then simply shifting the two images so that the object that you want on zero parallax aligns exactly works surprisingly well. On the downside the resulting picture is slightly smaller (you reduce the pixels covered by both the left and right eye picture because of the shift) and you have to render a slightly wider picture in the first place. It’s a little annoying though that you have to do that outside of Animate and for every shot (and ideally every screen size that you aim at).

Guess I still have a lot of experimenting to do… Thanks for your help! :slight_smile:

I might not have all the answers, but let’s see if I can help a bit. We’ve seen evidence of a few Stereo 3D projects done with our technology, but everyone of course has to experiment a little to see what will work for their project.

1. There doesn’t exist something like this in Animate Pro. Since objects are usually by definition flat, it’s the way that you construct your scene that defines where the parallax is. I’d recommend doing some tests, with dynamic camera movements and static ones, for the initial scene construction. Then you can create a template to reuse in other scenes as a basic guide.

2. I wonder if you have too much convergence set here on your cameras. Too much convergence gives the keystoning problem outlined here:

http://www.sky.com/shop/export/sites/www.sky.com/shop/__PDF/3D/Basic_Principles_of_Stereoscopic_3D_v1.pdf

If you’re really concerned about it, you can render out the images with no convergence then put the convergence in post. I haven’t seen this tried but I imagine if you can do this with real cameras, why not digital ones?

3. You can use the Ortholock module to keep layers always facing the camera. This may help.

4. Ah, like I wrote to 2. Indeed, if you do everything parallel, then you have everything appearing in front. But you can adjust this in post, I believe.

5. I would check out the link I listed above, which lists some common problems with 3D. I thought it was quite informative.

~Lilly

No worries! Good luck with your experimentations, and I hope everything turns out well!

~Lilly