Jump to content

Chroma Key Work


Recommended Posts

Hello. I am for lack of a better word, a newbie, when it comes to VFX. I'm also disadvantaged as I am an economics major, and have never studied filmmaking in any capacity. However, I have for the past 1 1/2 years, studied cinematography from any source possible. (mostly books). I understand the basic concept of chroma key but I am confused about a couple of things, that I was hoping could be cleared up on this forum.

 

1) When should I use Blue vs. Green in a chroma keying situation. I heard something about blue being further away from red on a vectorscope.

 

2) When the desired shot is a focus pull to an element that will be in the matte, is it better to shoot the footage with a focus pull, or save it for post. Does the lens create a blur around the edge at some point in the transition, making it difficult for an accurate key?

 

3) In many behind the scenes, where I see professional VFX being done, I notice these green x's of tape on the background. I haven't found any literature online about them. My guess is that they are reference marks for the compositors. Could anyone validate this for me?

 

4) Are there any professional methods to pulling the "perfect" key? My understanding comes from having the actor as far away as possible from the blue or green, because the further away, the less spill. Also having the background and the subject separately lit. From my understanding it is usually common practice (depending on the nature of the intended composite) to light the subject in tungsten, or anything with a 3200K colour (yes, i'm canadian) temperature, and the background with fluorescent, or green filtered light (in this case assuming the chromakey screen is green). This allows the colours to be more separated in the vectorscope allowing for a cleaner key. Are there any other things I would need to know in order to pull a "professional" key.

 

4b) I guess what I want to know in general, is what separates your average chromakey work from the absolutely astounding?

 

This is my first post, and I feel it warranted a small intro. Sorry for it's length.

 

Thanks in advance,

Michael Sousa

Link to comment
Share on other sites

1) When should I use Blue vs. Green in a chroma keying situation. I heard something about blue being further away from red on a vectorscope.

It depends on what you're shooting. If your subject is wearing blue jeans, you'd probably want a green screen. If your actress has flowing blonde hair, it's much easier to key against blue, whereas brown hair is easier to key from green. There are some other factors as well- blue is typically the grainiest emulsion layer in film, which can make keying blue more difficult. Lower-end and prosumer video cameras frequently toss out more blue information than green, although there are disagreements over whether or not this actually makes a difference. You might also want to take into account the environment your subject is going to be placed into- if he's going against a blue sky ultimately, the spill light from a blue screen already looks more natural, and can be corrected more easily, although you can still make it work with green.

 

2) When the desired shot is a focus pull to an element that will be in the matte, is it better to shoot the footage with a focus pull, or save it for post. Does the lens create a blur around the edge at some point in the transition, making it difficult for an accurate key?
In general it's best to get as much in camera as possible. Pulling focus on a lens does more than just change the blur- there's also a subtle shift in focal length, and changes in various lens artifacts, such as distortion. Keying defocused stuff- especially loose hair- can be really difficult, but most of the time you've got no choice. If you've got a specific reason to do it in post, then go for it, but you should typically aim to keep it in-camera.

 

3) In many behind the scenes, where I see professional VFX being done, I notice these green x's of tape on the background. I haven't found any literature online about them. My guess is that they are reference marks for the compositors. Could anyone validate this for me?
Tracking marks. All of the elments in your composite need to look like they were shot at the same time, so if the camera is moving around your subject, the background needs to move in exactly the same way. Tracking marks allow compositing artists to figure out how the camera was moving. Sometimes you'll see marks on C-stands or whatnot in the middle of nowhere- these are for parallax to be used in 3d camera solving, which allows you to determine the exact 3d movements of the camera, and replicate them to "photograph" CG objects that need to be inserted into the scene.

 

4) Are there any professional methods to pulling the "perfect" key? My understanding comes from having the actor as far away as possible from the blue or green, because the further away, the less spill. Also having the background and the subject separately lit. From my understanding it is usually common practice (depending on the nature of the intended composite) to light the subject in tungsten, or anything with a 3200K colour (yes, i'm canadian) temperature, and the background with fluorescent, or green filtered light (in this case assuming the chromakey screen is green). This allows the colours to be more separated in the vectorscope allowing for a cleaner key. Are there any other things I would need to know in order to pull a "professional" key.
Color temp should be whatever the DP would normally use to light the scene. It's ultimately irrelevant to your composite. You can use whatever lights you want for both the screen and the subject, although super-green Kinos help to really saturate the green. The most important thing [which never actually happens, but you want to try] is that the screen is lit evenly. Generally try to aim for 1 stop under key, although this can vary for all sorts of reasons. Distance is important as well, to minimize spill.

 

4b) I guess what I want to know in general, is what separates your average chromakey work from the absolutely astounding?
Probably the lighting. It's absolutely vital to match the quality of light as closely as possible between the subject and the environment. If you've got someone standing in what's supposed to be a bright sunny day, and you've got really soft, diffused lighting, it's just going to look wrong.

 

As far as the compositing, the edges are what's most vital. It's easy to use a keyer to get a "one-touch" result, but it's never really perfect. You've got to refine it carefully, and use several different keys blended together most of the time. For instance, someone's hair will probably take a different key than their waist.

 

Check out www.fxphd.com, they've got classes in all of this stuff, and I've found it to be really really helpful, and totally worth the money.

You can also check out The Art and Science of Digital Compositing, and The Encyclopedia of Visual Effects

Link to comment
Share on other sites

Thank you Scott for that informative post.

 

1) Is their a way to know which colour is best regarding which colours work best with which screens? I did not know that about brown vs. blonde hair. Or is it simply experience?

 

3) Could you elaborate on the tracking marks? How does it relate their movements in 3d space? Is their a way to find out its movement in the z-axis by looking at the change in size? Is it by eye? or are their ways to exactly measure it?

 

4) Also, you mentioned keeping the screen one stop under the key. Do you mean the subject? Or just the key light on the subject.

 

Again thanks for the detailed response. I'm quite happy about discovering this forum!

Link to comment
Share on other sites

1) Is their a way to know which colour is best regarding which colours work best with which screens? I did not know that about brown vs. blonde hair. Or is it simply experience?
There are particular things, like hair, that are known to be better with one color or another, but mostly it's just experience and looking to see what other people have done in the past and learning what you can from them. There are no real rules about it, it just depends on what you're shooting.

 

3) Could you elaborate on the tracking marks? How does it relate their movements in 3d space? Is their a way to find out its movement in the z-axis by looking at the change in size? Is it by eye? or are their ways to exactly measure it?
If you want to learn about this stuff, it's probably worth your while to check out some compositing software and learn how tracking works. Any program will be able to do 2d tracking. For 3d camera solving, you generally need a specialty application, such as MatchMover Pro or Boujou or PFTrack. What these programs do is take a bunch of 2d tracks and some information you supply them, compare their relative movements, and using the principles of parallax, determine where these points must exist in 3d space, and where the camera must have been. It's really cool technology, and quite complex, but worth knowing.

 

4) Also, you mentioned keeping the screen one stop under the key. Do you mean the subject? Or just the key light on the subject.
If you're exposing at f4, for instance, then the light on the screen should be f2.8, as rule of thumb.

 

I had also heard that shooting with a higher shutter speed will help reduce motion blur, and give cleaner edges. Is there any truth to this? Again sorry for the double post, and thanks for the reply.
Yes, that's correct. It also affects the aesthetics of the shot, however, so again unless you had a specific reason to, you wouldn't want to set your shutter angle differently for one particular shot than you had for the rest of the show.
Link to comment
Share on other sites

  • 1 month later...
In general it's best to get as much in camera as possible. Pulling focus on a lens does more than just change the blur- there's also a subtle shift in focal length, and changes in various lens artifacts, such as distortion. Keying defocused stuff- especially loose hair- can be really difficult, but most of the time you've got no choice. If you've got a specific reason to do it in post, then go for it, but you should typically aim to keep it in-camera.

 

I'm going to have to disagree on this one. If you can do the focus pull accurately in post, then I'd suggest going this route. If your foreground element is effectively planar - a single person or a row of people without anything in front or behind, then I'd really recommend keeping them in focus and doing the pull focus in post.

 

It comes down to which problem you'd rather have... Would you rather have a track that's slipping and is a pain to match because the out-of-focus tracking markers are coming into focus part-way through the shot, and, due to the apparent size of the marker, causing your track to shift. Or would you rather have to add a slight breathing to the foreground element as you add your focus pull later.

 

Obviously if there is a good deal of depth information in what's being shot in front of the screen, then you'll definitely want to do your focus pull in-camera, as trying to cut up a greenscreen shot to add that kind of pull in post is a huge amount of pain.

Link to comment
Share on other sites

  • 3 weeks later...
Nope. Only in specialty applications at this point.

 

If you do decide to sign up for fxphd.com, let me know so I can get a referral bonus ;)

 

True. But you can use those apps to export back a track into AE to use in your composite.

I am not a huge fan of AE's 3D implementation, but it does work.

Link to comment
Share on other sites

Have you played with Nuke at all? Nuke's 3D system is fantastic - especially for anything involving camera projection or cards in 3D... I can't remember off-hand which formats it supports input from - I'm pretty sure Boujou is one, and possibly Maya (amongst, almost certainly, others) - and, as it's all Python, it's easy enough to write your own importer for other ASCII formats...

Link to comment
Share on other sites

Have you played with Nuke at all? Nuke's 3D system is fantastic - especially for anything involving camera projection or cards in 3D... I can't remember off-hand which formats it supports input from - I'm pretty sure Boujou is one, and possibly Maya (amongst, almost certainly, others) - and, as it's all Python, it's easy enough to write your own importer for other ASCII formats...

 

It's great.

And expensive.

Link to comment
Share on other sites

  • 2 years later...

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...