Things I Didn’t Learn in School Pt 2: Why Does My Greenscreen Look Like #&$%??
My attempt at a regular blog series hasn’t doesn’t much for keeping me blogging regularly, has it? I really suck.
Admittedly, I started to write about aspect ratio a couple of weeks ago. I felt I had a firm handle on it and could easily explain the basics in a blog. Then I opened up some sites to do research to verify claims. Then I started reading more, opening more links, reading more. Then I got really REALLY confused and command-q’d everything and went outside.
I think I’m ready to come back at it after that little break. But not aspect ratio, that’ll take a bit more time. That crap is complex.
I’m going to be shooting on greenscreen for the first time in a million years soon, so I thought it’d be a good time to cover this one..
Crap I Didn’t Learn in School Pt 2: Why does my greenscreen look like #*^$???
How different formats capture color and how it ruins your life.
When I was in school, I had to do a project with green screen. The pictures in this post are from shooting that project. It was the first time I’d ever really worked with it. The project was to take a short scene from a film, shoot something against a green screen to insert into the shot, and combine it all so it makes sense. We were shown how to pull a key using Final Cut Pro. My school had a fairly nice greenscreen lab to work with. It had issues though. The lighting was iffy and didn’t always work. It was also theatrical lighting from the ceiling, which isn’t ideal. The room was somewhat small, making our running shots difficult. The cameras we had to shoot this on were Canon Gl2’s (standard definition, on MiniDV tape). We chose a scene from Lord of the Rings.
It did not turn out very well. In fact, it ended up looking like total crap. For one reason or another, we ended up having to do the keying in Adobe Premiere instead of Final Cut Pro. After we presented it to the class, our teacher tore it apart and told us how awful we were. I didn’t feel too bad about it, because everyone else had the same issues as we did. Total crap. Our end result was so hilariously bad, in fact, that I left it on YouTube and it has over 30,000 views because people love to make fun of it.
Since then, I’ve gone on to do a lot more research about color keying and realized this project completely set us up for failure from the beginning. It could have been a good opportunity for us to learn these things by seeing our failures come to life before our young innocent eyes, but it didn’t go down like that unfortunately. I’ve gone on to do quite a lot of keying professionally. I was given a large amount of material shot by other people before I came to my current job, and it was shot on MiniDV and then transferred to another MiniDV, then given to me to key. After you read this, you’ll understand why those keys took me a while to get to an acceptable level.
Here’s what we should have been taught:
The greenscreen you’re shooting on has to be evenly and brightly lit. We didn’t have any problems with this, but a lot of people don’t really grasp that it has to be EVEN. Like, EVEN EVEN EVEN. Test it out before you shoot. If you have a huge difference between greens, you’re going to have problems. One way you can check this is by turning zebra stripes on on your camera. If you can’t get the lighting right, you’re going to be screwed no matter what you do in post.
One reason we were extremely doomed from the beginning is that we were shooting on MiniDV. DV has a color space of 4:1:1.
4:1:1? What the hell is that? Yea, I know. Nobody said there would be math involved. Apologies.
To simplify it a lot, there’s this thing called chroma subsampling. It’s a way of compressing the video signal down – there is less information recorded for the chroma (color) than for the luma (brightness). It usually doesn’t have a big effect on how audiences perceived it because it’s fairly efficient. Unfortunately, when you get into heavy color correcting or color keying, you will seriously perceive it and then you will scream. There is a huge long explanation that could go here, but if you’re interested, you can Google it and spend a couple of days attempting to understand it all.
An analog video signal can be denoted as Y’CbCr. I could also go into this a LOT more because there is so much more to this and it’s not always Y’CbCr and when it’s digital it’s YUV and blablahblah, but for simplicity I’m just going to leave it at that because it’s not the point of this article). Y is luma, Cb and Cr are chroma. As you can see, 3 parts, like 4:1:1.
Full color space (no chroma subsampling) is 4:4:4. That means basically full luma, full chroma (blue difference), full chroma (red difference). There is green mostly in the luma, and a little bit in the red and blue.
So if you have 4:2:2, that means full luma, and half as much chroma.
4:1:1 then means full luma, 25% chroma.
There’s also 4:2:0, full luma, 25% of the total chroma in a different way.
What this means to you as the person attempting to pull a key: depending on what format you shoot, you might be simultaneously shooting yourself in the foot by throwing away tons of valuable color information.
So the MiniDV we shot on was 4:1:1. As you can see, we have thrown away a large amount of our color information. And did I mention we had our talent running at full speed? So we’re recording rapidly changing visuals with only a fraction of information. Oh man. So smart. And shooting on HDV would have been no better. You would think something with an “H” in it would be awesome. HDV is actually not awesome in a lot of ways. And HDV is 4:2:0. While it’s possible to get an OK key from DV or HDV, it’s never going to look really good, and it’s going to be a lot of work.
And of course, DV is interlaced, so our character’s motion would look like crap. The interlacing and lack of full color information made the edges look like such utter crap, it was impossible to smooth it and still keep it looking natural.
And now you might see why recording a greenscreen shoot on MiniDV then copying it to another MiniDV for editing might be one of the worst things in the entire world.
Another nail in our coffin: Final Cut Pro for keying. FCP is abysmal for keying, especially keys that are going to take a lot of effort. Premiere is not much better. But we were taught in FCP, so we thought it was ideal. What we should have used? After Effects. Keylight is a powerful keying plugin that comes with After Effects, and it has so much more flexibility and power when it comes to layering, masking, and spill suppressing. Plus, we were trying to insert our shot into an existing shot we had no input in shooting. We could have used a light wrap to fit our character into the scene better.
And because we had no color information to work with, and were trying hard to match our color to that of the scene after we finally had something of a key going. We were effectively color grading a pile of mushy crap that had all the life sucked out of it by the source tape and multiple instances of a crappy key.
I never learned a single thing about RGB/Y’CbBr/YUV/chroma subsampling stuff in school. It would have been extremely useful to understand at a basic level why different kinds of video are actually different, to put it under the microscope. I realize this is pretty technical, specific stuff, but wouldn’t half a class going over that be more effective than spending half a class telling us we suck?Alas..
Now that you know there’s a lot more going on in your files and tapes than you originally thought, I highly recommend you search for some more complex articles on the subject. A decent place to start: Wikipedia. There are people that have done tests and demonstrations to show color information going “poof” right before your sad little eyes, and it’s quite interesting.
Summary: Don’t shoot DV and expect a good key. Seriously.
OK, I spent the whole post talking about my horrible awful piece of crap greenscreen project. I guess it’s only fair that I link it now. Remember: this was a long time and many many keys ago and does in no way reflect my ability now! 😀 And yea, it probably would have helped if we had dressed her in the appropriate clothing or edited it correctly at all, but no. Just know it was a group failure, and ours was probably one of the best! I would embed it here, but since it “matched third party content” according to YouTube, embedding is disabled. You’ll have to go to YouTube to view it.
By the way, that kind of sucks that YouTube does that without regard to fair use in education. Obviously this exists because I did it for class. I could probably file a claim against YouTube and ask them to take this flag away from my video, but when you do that it looks very scary and intimidating. And really, do I care if this particular video can remain online at all? Not really..