mutiple render contexts and camera

    Feb 11 2006 | 10:44 pm
    Hi list,
    A few weeks ago we investigated the best way to use multiple fullscreen outputs.
    The best way to deal with this is using multiple context doing the same things in stead of one context and switching renders.
    When using this some other fine tuning is aquired when wanting to transform in the z axis. I prepared a patch for testing and maybe posting solutions.
    I already tried lookat & lens_angle but I couldn't get my mind around it that way.
    In the example patch I moved the second render with -2.15 0. 0. so logically I thought the camera needs to look at -2.15 0. 2. (Actually, just moving works fine when not using the z)
    When I however turn in z using the handles, the screens do not align anymore. Does anybody has any hints how to counter that?

    • Feb 12 2006 | 10:34 pm
      Little update,
      It is offcourse obvious that this works in this way.
      Imagine you got this view from above:
      X / / / --.--.--
      Where "."'s are camera's with an angle of 45 degrees "/" and "" indicates where the viewing edges are of those camera's
      X marks the crossing point of those lines. Then all behaviour is explained, what i however want to reach is this | | | | | | --.--.--
      Where the two camera's look straight ahead and do not cross each others path of viewing.
      Is there a way to accomplish this?
    • Feb 12 2006 | 10:49 pm
      Hi Maarten,
      All I can say is you have to transform the position and rotation of one camera in relation to the other. I started a tread about stuff like this because basically want to do the same, but with an object in relation to the camera. I started working on it, but got stuck and haven't found the time and courage to get my head around it since then. You can check out the topic here. 38&r id=0&srch=3d+coordinate#msg_61938
      hth, grt T_
    • Feb 12 2006 | 10:51 pm
      There's a type of lens out there that doesn't cary depth information in the final image. I think it might do something close to what you describe, but I can't for the life of me remember what that type of lens was called. I tried searching for othogonal lens and didn't turn anything up. I believe this type of lens is used by computer vision people most often because of the unique properties of the image.
      Of course, I could be constructing a memory from nothing and this could all be lies. But for some reason, I have this idea of a type of lens that distorts the image in such a way that apparent depth is lost.
    • Feb 12 2006 | 10:53 pm
      > what i however want to reach is this > | | | > | | | > --.--.--
      i' have not tried this, but what your are suggesting seems to be orthographic. see the @ortho attribute of
      another thought, what if you had two cameras with angles of 22.5 each with their adjacent edges/clipping planes aligned.. something like:
      | / | / | / |/ ----.----
      just brainstorming,, :)
    • Feb 12 2006 | 10:59 pm
      Whoops, I thought you were talking about real cameras.
    • Feb 12 2006 | 11:01 pm
      i'm using the command ortho 0/1/2 for an orthogonal viewport projection. 0=off, 1/2=different versions of ortho, i forgot what exactly.
      best, d.
      _____ dominik busch _____
    • Feb 12 2006 | 11:05 pm
      ps ortho applies to the render object d.
      _____ dominik busch _____
    • Feb 12 2006 | 11:17 pm
      I just tried Ortho, forgot the function existed. :)
      it seems to get rid of some of the cludder or messiness but it doesn't align properly. Maybe using Ortho with some rotation or camera functions will do the trick.
      Thijs thnx for the link, I will look into it.
      Thnx for pointing out ortho and for getting me some steps closer.
    • Feb 12 2006 | 11:19 pm
      In reply to Jonathan placing them in one point hint.
      I'm actually not using two but more camera's if you get more you will eventually make circle I think.
    • Feb 12 2006 | 11:43 pm
      Well I made it work with the example patch in my first post. However I offcourse loose the z axis in the process. So I will have to do some testing how to solve that when using multiple objects. I guess I will have use scaling instead of z positioning
      It will do for the given patch, it is however a simplification of a more complex patch. Thnx for pointing out Ortho, I forgot it was there.