I’m trying to match a skinned, posed 3d character to a background photo of a figure. This is for compositing of several passes of 3d elements onto the photo model’s skin for hi rez print output.
Do do the most precise job, I need a way to zoom into a region of the camera viewport without altering the camera – like zooming in the frame buffer, but in the viewport, where the wireframe lines stay their normal width and the geometry remains clear.
The simplest thing would be to get a screen grab of the viewport at a much higher resolution than is displayed. This only meets halfway though, because I’d like to be able to do final mesh tweaks interactively while zoomed in on a region of the camera view.
I just started investigating the camera correction modifier. Trucking a duplicate camera close to the model, and adjusting the perspective to match the main cam … I’m not enough of a mathmagician to know if this is the right approach.
=======================================================
further details about the job, which help explain why I need this.
-Using a single 23 inch apple monitor that is not designed to rotate into portrait mode
-There are 40 shots of the figure which need to have elements comped onto them.
-Focal lengths, camera position, etc are widely varied among the shots. Nobody bothered to take down this sort of info at the photoshoot.
-The photos are being retouched before they come to us, warping the subject into often anatomically incorrect, idealistic shapes. This is done “artistically” per pose, and inconsistent from shot to shot.
-many fine details which need to follow the body contours. Some are only 3 or 4 pixels wide even with our 15,000 px renders, so there’s not much room for warping the render afterwards.
-I’ve tried all sorts of combinations of see-thru, transparent materials, backface culling, opacity mapped planes with the photoplate aligned to camera… but max vport transparency is so lousy I can’t get the clarity I need to fine tune the mesh.
well, after exploring graphics window methods in maxscript documentation I found out about the virtual viewport, which only woks in openGL. I guess I’ll use that for now. I never did know what Virtual Viewport was, as I’ve always used DirectX driver due to it’s vastly superior performance.
Just wait till max 2008, there will be several maxscript (and SDK methods) available that should meet your needs.
I am definitely looking forward to Max 2008, though I’ll be done with this tediousness by the time that comes around. I really would like some sort of Virtual Viewport under DirectX.
I couldn’t deal with using OpenGL after a few hours. 75,000 polycount onscreen and I can barely rotate the viewport… This on a Four Core 3.6ghz Xeon Boxx with QuadroFX 4500. That just ain’t right.
Hi yoshimitsu,
This script: ImagePlane13.ms from www.scriptspot.com could be dealing with part of your problem.
It creates a plane normal to the camera to use as background plate.
You could enhance it, linking the UV coordinates of that plane to the camera FOV angle and position.
When the camera dolly (distance from camera to plane) or FOV angle changes , the UV coords. updates accordingly.
The math involved is simple: 1/2 tan FOV = (1/2 width plane) / (distance from camera to plane)
Fernando