compositing


(Nayman) #1

I am having trouble getting a cgi object to go behind a real object.

I understand alpha masking, but that onlly works for stills.

In my spider man movie, i have lots of CG interacting with real life, and i need to know how i can cover my CGI character with parts of the animated background.

I have tried rotoscoping a new layer ontop of it in photosop, butit doesnt line up for some reason…

I need this resolved quickly, so any oppinions would be welcome.

thanx


(d52477001) #2

yes i have the same problem. i have a video with someone moving and i need to have something cg behind it but i can’t get the object to be behind the moving person accurately.


(rwenzlaff) #3

Why do you say that? You need to make an animated mask, but you can still use the technique.

Rotoscope a(n) object(s) the shape of your forground objects (they can just be planes if their outline doesnt change). Animate them so they match the motion of your live objects. Use a texture with the “ENV” button clicked. Place your CGI object in this scene. The mask will render with the same with the Alpha as your background so when you composite, your CGI object will be hidden by the live footage when it goes behind the mask.

Because of the lack of depth in the 2D footage, your mask really only needs to be accurate at the points/times the CGI object will cross it. The eye wont be able to tell is the live footage is renderd on the mask or the background, unless the CGI object breaks the “cut”.

It’s a pain, but it’s really the only way to do it. “Going behind” is inhearently a 3D phenomenon, and all the Z info is all lost when the live action went to 2D film. It’s the only reason matte paintings work…

Bob
www.soylent-green.com


(rwenzlaff) #4

Also, I forgot to mention, you can do this all in one pass by using a plane for the background and using the live footage as an animated texture. Just use the “WIN” mapping for both the background plane and the mask. It will use the XY coords of the camera view to map the texture, so it should laways line up…

Bob


(harkyman) #5

That last trick would work, but Blender would interpolate the original video footage, and you’d quite possible lose resolution/quality. I’d go with the animated mask and do my composite in a separate step.


(rwenzlaff) #6

As long as the render output size/aspect ratio is equal to the live input size it should be a straight 1:1 mapping (and you can turn interpolation off in the texture settings).

If the output size differs from the imput size, the output is going to be interpolated even in the sequencer.

Bob
www.soylent-green.com


(kattkieru) #7

Do some research on the web about Chroma Keying. It’s a simple effect that a lot of movies use to replace skies, etc, without having to go in frame by frame and select the area. It does this by removing a specific color (or a specific range of colors, like the tolerance of the Photoshop wand) and replacing it with parts of another movie.

It sounds to me like you’ll need to chroma key a wall or sky or something (IE, spiderman sneaks up behind a lady), and have it replace colors with your CG footage. I assume there is a way to do chroma key with alpha on the replacement stream, allowing you to only add spiderman to the scene without the black background around him.

http://heroinewarrior.com/ has a program, Cinelera, which does all this and more, and is free. Unfortunately it’s only for linux, but serious Blenderheads aren’t using Win32 anyway. ^.^ On Mac, I think there are plugins for iMovie (may cost you 20$), and I know Final Cut Pro 3 has this functionality built-in.

The method described by other users above will work, but it requires more effort and IMHO should only be used as a last resort, or as a replacement if you can’t find software for compositing.