View Full Version : Seamless Compositing

Dipesh (India)
08 August 2002, 03:48 PM
How do the CG animated 3d characters get seamlesly blended with live action movie! Does anybody know what the process is in detail!
I only thing i know is that chacracter are shoted againts blue screen than composited but certainly there would be hidden process like cinematography, camera tracking etc.
Any one knows in details.

08 August 2002, 09:24 PM
Your question is way too broad. It's like asking someone how to build a car, engine and all from scratch. I suggest you go out and buy the book "The Art and Science of Digital Compositing" by Ron Brinkman. Also check out some of the tutorials that come with the compositing software you are using.

Dipesh (India)
08 August 2002, 10:31 AM
is there any site which explains the procees in detail!

08 August 2002, 07:43 PM
here's a basic rundown of the process, and I will assume you know nothing about it so forgive any "condescending" info:

The 3D animation/object is almost always rendered as a tiff or targa (.tga) file or sequence. Targa files support an "alpha channel," which is the mask of your object. This means that where there is no object the image file will be invisible. Most other file formats do not support this, and it is the key to good, clean composites.
Sometimes the sequence is rendered with "z-depth" channel info, which stores depth information about the file. The rendering engine usually renders this out as a separate file, and you would use it in your compositing package to set depth of field based on distance from the camera or other effects as needed. Using z-depth for depth of field, rather than letting your rendering engine perform this task, is pretty common because you can adjust the precise amount of blur as needed, frame by frame, and it usually looks cleaner and more true to life than the rendering engine can do, ESPECIALLY when compositing with foreground elements that need to be in sharp focus. This is an advanced issue, and I recommend avoiding it for now.

In almost all professional situations the 3D element is rendered out in multiple passes, such as the highlight pass, shadow pass, beauty pass etc. The beauty pass is the actual object with ambient lighting and no reflections, highlights or shadows. Maya and other high end 3D packages allow you to render in passes, but usually you have to manually assign flat black shaders to the objects to get your specular highlight pass etc. This is time consuming and a little difficult at first, but well worth the time saved in the long run. These individual targa sequences are then layered in the appropriate stacking order (beauty pass on bottom, highlight on top etc.) in your composite package to create the final image.

In your compositing program, the alpha channel is used as the "matte" for the object. This will leave the object 100% opaque and drop out the background.

Doing all of this usually looks good for the first composite pass, and then it is fine tuned for effect. This is where rendering in passes really comes in handy. One example of this is that you can increase the highlight or change its color, if the art director so chooses, by simply manipulating the "highlight pass" layer in your composite. The alternative is having to re-render the entire scene, which is a total waste of time and will eventually be repeated many times until you are happy with the results. This would be a time crunch disaster in a production environment.

Not being in a professional situation, I would suggest starting out by rendering your animation/image as a targa file, which allows for alpha channel (make sure to specify "render with alpha"), and bring it into any compositing program. Set the layer with your 3D object to use the alpha channel as the matte (this is done automatically in After Effects when you import the image), drop in a background image and viola! Your first composite.

Good luck!

08 August 2002, 07:02 AM
I agree with beenyweenies.

Aside from the above mentioned techniques, aproaches that would make a cg character blen into the scene seamlessly is to simulate/approximate the lighting condition/setup of the live plate where the cg images will be composited into. This is the duty of the lighting technical director (and his crew). One of the technique that would aid them is to place lighting references in the actual shoot plus the mock-up/dummy/marquette of the cg character. The matchmover will also have to match the camera movement in the live plate to the cg to make it convincing that the cg character is indeed part of the scene.

Those marquettes I mentioned will also guide the compositor to color correct the cg material as close as possible to the dummy guide shot for reference plus his input to what he think is the best color correction using the elements given to him by the 3d dept.

The compositor will then add additional elements like shadow, smoke, depth of field treatment, etc. for finishing.

This may sound easy but it's more than that in actuality.

Dipesh (India)
08 August 2002, 09:00 AM
oh my god!
that was really a great piece of technical information from beenyweenies & wireframe! you have cleared many of my doubts!
once again thanks!
more information is invited!

CGTalk Moderation
01 January 2006, 03:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.