Converting 2D Video to 3D Stereoscopic Video (NOT Animation to 3D)

Hi, I just want to know if it is possible to convert an Actual 2D Video shot with a Cinema Camera (or any camera), to Stereoscopic 3D (or 3D for that matter) within Blender?
If yes, how can one start the process of doing it? What are the basic steps to accomplish that? If the subject is too broad to explain here, can anyone give a breakdown of the steps that should be followed? If that is also too much to ask, at least a Link or a Forum that handles this workflow.
I have already done Google and YouTube searches, but all I could find where the methods to obtain an Animation in 3D Format. That is not what I am looking for, at least, not for now.
I assumed this topic would belong in the “Compositing and Post Processing” section. Feel free to correct me.
I am a Freelance 3D Artist in Fiverr:
Feel free to contact me there, if required. I have only been learning Blender for the past 3 months. So there are many areas in Blender that I am unfamiliar with. Forgive me if this is question is too trivial (for the record, I believe it is not). I think if Blender already has this feature, that would be a huge deal. It is already an intuitive and powerful software. This would make it epic. If this is an unexplored area, then I thought it might be helpful for a lot of people out there.

1 Like

Stereo conversion basically consists of:

  • breaking your image elements down to different depth layers
  • separating them in depth
  • rendering two new views from your scene.

Simple, right. For separation you can use manual roto work, automated methods or a hybrid of both. Usually for simplistic scenes depth/object detection can work but characters are mostly rotoed if you need correct edges etc. One very labor intensive phase is inpainting, that is creating data for areas that are occluded in original camera view but become visible in other view.

Depth separation can, again be done either automatically, using depth generator tools, depth maps etc or you can manually assign depth values to different elements by either creating a depth map or laying them out in 3D space using 2.5D methods. What you also need to take care of is the “roundness” of objects, usually characters, because otherwise you get flat, card-like effect. You can also project images to geometry to get proper shape representation.

New views are usually generated for two new synthetic viewports, not original + one new, because synthetic views are more similar to each other in edge qualities etc.

Thanks for the reply. I haven’t been active here for a long time. Your answer is indeed helpful. Thanks for spending the time. Have a good one.