3 Replies Latest reply on Jun 21, 2014 6:46 AM by feebus789

    3d rotoscoping?


      Question: Is it possible to get the pixels of an 3D object from inside a footage (gaming video to be specific) and convert it into a 3D model (.obj or .c4d file)?


      For example,


      In game: the sky, walls, etc. are all white except my weapon sticking in air and its position is frozen (always fixed.)

      This weapon is the only thing I can see, nothing else.

      I can walk to it and see it closely.


      I assume that.... I record and while recording, I would circle around the weapon (160 angle) and then stop record.

      Then I want to convert that weapon (from the recorded footage) into a 3D object somehow. Problem is I don't know what or how to do that...

      I have an experience in tracking the pixels (track and trace, i.e. putting a 3D text onto a tracked point), that's it.


      Any advice / tips would be helpful. Thanks

        • 1. Re: 3d rotoscoping?
          Rick Gerard Adobe Community Professional & MVP

          You cannot create a 3D model from a 2D video in After Effects. What you are asking is very difficult to do automatically with any kind of software. The easiest way to recreate your object in 3D is to find a side and top view of the object, grab some still frames and use them as a reference to build the model in a modeling app like Blender (free) or C4D (not the Light version that comes with AE).

          • 2. Re: 3d rotoscoping?
            Mylenium Most Valuable Participant

            It can't be done simply because there is no such app or feature. Aside from a few SIGGRAPH research papers there simply is nothing. All photogrammetry tools/ 3D reconstruction tools otherwise operate on photos and require manual work... sorry kid, you're 20 years too early with your idea...



            • 3. Re: 3d rotoscoping?
              feebus789 Level 1

              Oh well, but thanks. That's what I was thinking. Finding sides and use em to build the model although that won't be perfect but it will work just fine. There's an alternative, which involves in hacking the file (where the model is located) and extract it, but that requires a whole different level of ball game: what if it's compressed? What if it's encrypted? Afterwards, I guess it's all about learning the file format and structure and then map the data at certain offset (s) and whatever else in order to extract the 3D model. Which can be challenging since a lot of video games use different file formats and structures.

              And to the other poster, your last sentence was a bit offensive. So, no thanks to you.