QUESTION: I’m trying to achieve an effect similar to the recent Green Lantern movie, where the actor would be wearing a chroma key suit and I would then key it out and track the effect to it. Could a regular chroma key suit work? Would I need to place tracking markers of some sort on it?

First, there is a huge misconception that if you put an actor in a green or blue suit that you can then press a magic keying button and make him disappear - nothing could be further from the truth.

It is useful to have the actor in a monotone blue or green (or red, or black depending) suit to make auto-tracking masks easier, IF auto-tracked masks will be needed. A color key type suit can be helpful if the actor will be wearing "partials" (partial costumes or props) that will still appear in the scene. And even in those cases it is often best to use a suit close to the color of the final element,

What is more important is the CLEAN PLATE that is the scene behind the actor — how are you going to capture that? Either your camera is locked down, or you are using a motion control rig with repeatable moves. OR your are going to shoot only the plate, and then get the motion from a mocap stage.

And what level of motion capture do you need? Do you need it to drive a fully articulated "robot" 3D character?

Which method is best is hard to say, since you have not really told us anything about the shot.

So I will make some assumptions - You don't want to see the actor at all, only the ""3D robot" against the background. With this assumption, here is one possible workflow:

  • Shoot the plate for the shot where you want the action to occur. You should have the actor there for some rehearsal takes to nail down the camera move you want, but shoot it with the actor NOT in the shot at all. This could he hand held, on a dolly, or however you like - motion control is not needed, but a stop watch and someone calling out "key points" to match the rehearsals would be useful. (You should ideally shoot some of the rehearsals as a reference).
  • Take accurate measurements of some of the prominent features of the environment (exact size of door frames, etc) as this will be useful is setting scale in the 3D tracking pass.
  • Also, and this is critical, take multiple HDR "light probe" images of the scene of the clean plate. These will be used to bring the lighting not he scene into the 3D environment to be created.
  • Now take your clean plate, and do a complete 3D motion track using PFTrack, BouJou, SynthEyes, etc. From this clean plate you can determine and capture the entire 3D "environment", including the camera and all of its moves — even the focal length of the lens can be determined.
  • Then go to a mobcap stage to shoot the motion — essentially put the actor in a skin-tight suit, covered with very high contrast dots, and shoot her with multiple cameras at the same time (from different angles). It is important that these cameras are all "genlocked" - i.e. the shutters are all "in sync: with each other, all opening and closing at the same time. (Note: this was originally written a decade ago — mocap technology is much more advanced today).
  • If necessarily, add elements on the stage that match physical properties of the background clean plate (a chair, a railing to jump over, etc.).
  • Load this footage into a mocap program and use that motion mesh to develop your 3D objects and animation. You will bring this mocap animation into the 3D environment created from the clean plate.
  • Texture and finesse the animation of the 3D objects. Use the light probe to bring in the lighting environment(it is important the light prob be aligned with the orientation of the clean plate relative to the 3D environment).
  • Render out the objects/robots that will be added to the scene. You only need to render the actual 3D elements you are going to use — the rest of the 3D environment will simply be black. An alpha channel will provide the transparency information.
  • In many cases, you'll want to render out "channels", in other words, not just a final RGBA image of the 3D objects as they will be seen, but independent channels that will be combined in the composite - shadow, grunge, normals, etc.
  • After rendering, the 3D elements are dropped onto the clean plate and adjusted to match creating the final composite.

That is the *basic* run down - in practice, it usually takes a team of people that are specialists in each phase. If you are planning on doing this all yourself, expect a learning curve for each phase of development.

If, on the other hand, all you want to do is put the actor into an "ironman" suit that fully encompasses his body, then you can just shoot the actor in a monotone skin tight suit covered with high contrast dots, and shoot him on your actual set. When you shoot him, you need to use your main camera for the plate, but you must *also* cover him with at least two additional "witness" cameras (in genlock/frame sync).

Then use something like PFtrack and extract the motion of the actor, as well as the 3D environment and the camera.

Build the "suit" in a 3D app, using the motion/mesh from PFTrack to animate, render and then simply composite over the actor. This obviously works best when the new 3D suit is larger than the actor.

A "green screen" suit is not needed — in fact should be avoided — in this case. A better option is a color very close to the final 3D suit, and with high contrast tracking dots. By making the mobcap suit close to the same color as the final, it will give natural radiosity, reflecting the color into the environment.

There will be an issue of matching shadows, as the shadows of the larger 3D objects will not necessarily blend with the shadows cast by the actual actor. But if the shadows are kept very soft, they may even play without using shadows from the virtual 3D environment. 

And once again, it is very important to get HDR Light Probes of the scene.

Whenever you can have something real in the real environment to interact with the light and colors of the scene, the better off you'll be down the road. But of course all this needs to be tested and planned well ahead of the actual shoot.

This article was originally written circa 2010, but updated July of 2018

 

Advertisement from Google AdSense

Help | Legal | Contact Us

Copyright © 2012-2018. All Rights Reserved.