Stereoscopic Compositing destroys rendering

Dear Blenderheads,

recently I came across a problem. Images tell more than a thousand words, so here is what’s happening:


(.jpg)


(.png)

I am using a blendswap-Model and the stereoscopic camera addon to create a stereoscopic scene. It’s weird that the right image is alright and the left one is “tiled”?! Compositing nodes translate and mix the images via the “screen” method. I changed it to “add” (for me, there is no difference, since it is combined with a black canvas anyway) but this didn’t help. I’ve attached the blend file here:
http://adam.kalisz.co/bmps_gold_268_stereo.blend

Do you have any idea what the problem might be? I don’t want to file a bug report too soon, that’s why I am asking here first :wink:

I’ve used the search but couldn’t find anything related.

Thank you guys!

Hi BlenderEi,

Post a little old now and you may have solved it but good to see someone else interested in stereoscopy! I am currently in the process of creating a python script that will make the compositing of stereoscoping much easier only have to work on one node tree. Will post updates once its more usable.

I havent found any issue with your scene. I dont get the tiling as you mentioned im wondering perhaps is it a graphics card related issue? Or version perhaps. Im running 2.7. interested to know incase i need to to take into account with my script.


Also when compositing unless i am wanting RGB pixels to blend with some sorta math (add, screen etc) i usually would use “mix” with the include alpha box checked. Then it takes the alpha into account and preserves rgb data (though as you say, its adding to black anyway)

Hi mafster,

Thank you for your answer. Now I am also quite a bit late :wink:
Did you take a look at the current (in my opinion best) script for stereoscopy: http://www.noeol.de/s3d/
This script also creates the nodes for you.
I like your idea in providing better nodes, but how does your idea differ from the current script? You should probably get in contact with the developer and see how you can contribute to his script.

Also, Dalai Felinto is currently developing his “multiview” branch, which brings stereoscopy natively into blender. Unfortunately he needs some more time to complete it (you might also want to help here, if you’re proficient in C++).

About the problem in my stereo rendering:
It’s gone! It seems to be fixed. Thanks to the developer who did this :wink:

Cheers!

Interesting. I have a question, because you’ve got me curious now - why is a plugin needed for stereoscopy? Couldn’t you just parent two cameras (left eye & right eye positions) to an Empty, have them track another Empty which will be the camera focus point, and then do a render pass for camera A, then camera B? Pardon my ignorance on the subject, seems you’d be the right person to ask.

Ofcourse that would work. But it is more convenient if you don’t have to render seperately, but just have the left and right camera as passes of the same render, right?

I see, so it is a matter of convenience then. Suspected that was the case but wanted to make sure I wasn’t missing something.

Sorry for my late answer Quantum Anomaly,

You are right. That’s exactly what I am doing right now. The plugin does exactly what you are describing, but with just some button clicks :wink: It then saves the combined stereorendering out as a separate file. Now, I hooked up the nodes differently to make blender save out two separate images.

The reason being: I am programming a renderfarm with my team at university and I’ve implemented support for rendering stereo3d with the renderfarm. As an artist you have to prepare your 3d scene to save out separate images for the left and right eye, so that the renderfarm can recognize them and send them back to you after a frame has been rendered.

This is my workaround for the issue from my initial post (which by the way now appears again on linux!). But I don’t have time to file bug reports atm, since we need to get that renderfarm working and some exams are still due. Still, this was giving me a headache yesterday as I was trying to create a good test scene for the renderfarm in 4K and stereo3d… …after pressing F12 and waiting for minutes blender even crashed on me… …surprisingly our renderfarm (which works by starting blender via the command line) was rendering the first 200 frames without any issues.

I’d fix this bug myself, but I’m still at the start of learning programming highly complex software. Blender is too difficult at the moment. But I’ll get there :slight_smile:

I anyone has some suggestions, please don’t hesitate to comment. Thank you!

Cheers!