Ned, not sure if I'm too late on this or not... I recently did one of these using an approach pretty close to your 2D one. I built an 8K texture map of the LEDs in the screen pixels and multed it into my screen content which was 1 or 2K pixels across. (Sorry , can't remember all the details.) I also added layers of screen imperfections (8x8 pixel-blocks, light leaking between LEDs, fake spec kicks on LED domes, refresh flicker...) to help it not feel like a single "trick." I then whacked it on a card in 3D and "photographed" with a pretty wide angle lens, rather close up. At the closest we were seeing only a quarter to a third of the screen. The multiple levels of detail definitely helped sell it.
Pre-rendering elements and careful positioning of crops, along with a switch to be able to turn all the fancy stuff for interactive work made it surprisingly useable. All of this is a long-winded way of saying, don't be afraid to push 2D as far as possible before you abandon it. In the back of my head I was worried about having to build geometry for the screen, but we never needed it. The shots all got finaled and looked just lovely in the theater. Keep it simple (and changeable) for as long as you can. -t On Tue, Jul 3, 2012 at 8:26 AM, Ned Wilson <[email protected]> wrote: > My goal here is to use a Nuke 3D Camera to shoot a card or an imported > OBJ/FBX. The card would be the screen of the Jumbotron. The camera needs to > have the ability to get very close to the card, sort of like you were > shooting a Jumbotron at a stadium standing two feet away from it. > > I have experimented with a couple of things with varying success. My first > approach was to use a series of cards, each with a higher resolution > texture map on it, and as I got closer to the actual surface of the screen, > I would dissolve between the lower and higher resolution texture maps. This > was a bit error prone as the dissolves were never seamlesss; they required > color corrections, and if you were looking, you could clearly see the seam > between the two. > > My second approach was to do a really high res texture map. I started with > 16k. This image is incredibly cumbersome and memory intensive but it works, > you can get the individual pixels up to about 16 pixels wide and then it > starts to fall apart. > > My third approach was to try a 2D gag, where you take the LED pattern, > tile it, and multiply it against the scaled up source image. This also > works well, except that it can only accommodate mostly lockoff shots. > > In thinking about the proper way to do this, I would model something in > Maya with the pattern, and then instance that small piece of geometry > across the surface of a card or sphere, retaining the UV space of the > sphere so that I can drive the color with a texture map. > > Unfortunately, this presents an issue in that if I want to change the > resolution of the card, I have to go back into Maya, re-instance the > geometry, and re-export the setup to Nuke. > > My question is two-fold. First, is there a way to instance geometry with > Nuke, possibly with particles? Second, is there a better way to accomplish > this effect that I am not even thinking of? > > Thanks! > > -n > > ______________________________**_________________ > Nuke-users mailing list > [email protected].**co.uk<[email protected]>, > http://forums.thefoundry.co.**uk/ <http://forums.thefoundry.co.uk/> > http://support.thefoundry.co.**uk/cgi-bin/mailman/listinfo/**nuke-users<http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users> >
_______________________________________________ Nuke-users mailing list [email protected], http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
