Blue Spike Tubes Good Or Bad?

Hello,

My VFX supervisor is telling me that blue spike tubes make a blue screen harder to key than regular tungsten tubes (we're shooting tungsten balance). He says his software doesn't like that spectrum. I don't understand why this would be the case. I normally check the vector scope to make sure the blue is legal and trust post can make a good composite. Is there something I'm missing?

Thanks for your help.

Sincerely,

D. Gregor Hagey, CSC
DOP
Toronto

> He says his software doesn't like that spectrum.

Odd. The only spectrum he'll get back is what is both in the light and what's on the screen. It's not like he'll get back something that's not blue. Is the problem that it's too pure, or too narrow in bandwidth?

Art Adams | DP
San Francisco Bay Area
http://www.artadamsdp.com

>>My VFX supervisor is telling me that blue spike tubes make a blue screen harder to key than regular >>tungsten tubes (we're shooting tungsten balance).

A properly lit screen works fine using blue tubes. The key is "properly lit".

The biggest mistake I see is over lit screens using blue or green tubes.

With a properly lit screen, I don't understand your VFX supervisor's
statement.

Bob Kertesz
BlueScreen LLC
Hollywood, California

DVFX discussion wrote:

>> My VFX supervisor is telling me that blue spike tubes make a blue screen harder to key...

Recently ran into the same issue for a driving scene.

We ended up on BlueScreen for less spill & contamination and less offensive in the chrome & gloss of the foreground.

I really pushed to do it with blue tubes, but the Compositor was against it, saying it would be far too much foreground contamination.

Kept pressing the issue, I knew that they'd get a better matte from a pure blue screen rather than a blue/gray one.
I promised to shoot the screen very dark, and it ended up being even darker than I thought would be best, but the Compositor ok'd it.

I think the main issue is that some let the blue screens get to 70% on a waveform and it’s way too much blue spilling everywhere.
Otherwise I don't think any Compositor would argue with purity of screen colour once they start dialling in a matte.

_______________
Mark Doering-Powell
LA based DP

I'm curious, from a VFX Sup's point of view, what a proper exposure level is for a chroma screen?

Andy Jarosz
AC
708.420.2639
Chicago, IL

Yep - separation is the key.

Unless the blue spike backdrop lights are near perfectly isolated from the foreground, including bounce from the ceiling etc. surroundings, the subject will be contaminated, making it hard to pull a great matte despite of the purity of the backdrop colour. With tungsten lights, the only contamination you get is from the light that's reflected from the blue screen.

And that's also why many compositors also tend to like relatively low backdrop light levels: any spill will reduce the quality of the matte, especially in fine details like hair. The camera cannot see a difference between semi-transparent hair and skin with blue spill - both can be recorded as the exact same colour. So when keying, either you will get opaque hair with rather hard edges, or you get transparent skin (with nice looking hair). But you can't get both without two separate keyer settings and some tedious frame by frame rotoscoping between them, a task which compositors rightfully hate...

One more thing - I always try to keep the backdrop area as small as possible, as the larger the blue screen, the more it will reflect blue light to the subject. I usually have some black Molton stage curtains at hand, that I use to cover the unnecessary portions of the backdrop. You only need to have blue behind the actors (and their shadows, if you do floor too). Everything else can be covered in black and then garbage matted away rather easily in post.

Eki Halkka, DP / Jack Of All Trades, Helsinki, Finland

***********************************
http://spacewhale.fi
http://www.kolumbus.fi/erkki.halkka
***********************************

>> My VFX supervisor is telling me that blue spike tubes make a blue screen harder to key...

I also don't understand this statement. However personally I prefer not to use collared tubes (green or blue) to light the chroma key, because there is a higher chance of contamination of your foreground with collared spill.
Also, it only works for cases where subjects don't come too close to the chroma key screen. How would you use the spiked tubes if you were doing a full body shot of an actor, seeing his/her feet touch the chroma key
surface?

I've spoken to a couple of VFX artists about using green tubes for lighting green screen, and in all cases they weren't too excited about the idea. My understanding is that they are 100% comfortable with using the green screen lit by same colour temperature as the rest of the set, and not having an exaggerated chroma. So lighting with green or blue tubes doesn't necessarily benefit their workflow in a huge way, however it is very likely to bring other problems - like increased spills and reflections and bounce back in the foreground.

Cheers,

Pasha Patriki csc
Toronto, Canada
Director Of Photography
Digital Intermediate Producer

Having shot a fair amount of greenscreen over the last two or three decades (although it seemed like we were doing mostly bluescreen in the early days), I typically light them at 40IRE and have never had a compositor complain, but that's just my own empirical experience. So yes, it would be great to hear VFX folks chime in about level from their perspective. Also, whether or not they've found the optimal light level to be the same with green or blue backgrounds.

I've always tried to be fastidious about maintaining that constant bg luminance, lighting the screen itself with angles oblique enough to alleviate as much reflectance/ foreground pollution as possible, and flagging between the screen and the subject (just out of frame, of course) to handle the rest.

Regarding the blue and green spike tube discussion, I've used plus green or CTB on tungsten lights in the past, particularly when necessity dictated shooting with a client's portable, ersatz blue or greenscreen (e.g. often some material lacking deep saturation). I just thought it would make life easier for the compositor and don't remember the additional gel ever causing a problem. These days, I don't see that problem as often, since even most of the inexpensive portable screens seem to be more saturated.

Finally, as long as the VFX guys are going to weigh in, would you give us your impressions about using straw gel (or similar) on a foreground subject's backlight for further separation when doing bluescreen? I've often found that very effective for yielding more separation, not to mention pulling more believable keys when composited against sunny daylight plates. Of course with green screens, similarly using minus green on the backlights isn't usually an option (unless used in very specific situations- e.g. sci-fi, night time neon’s, etc.)

Thanks.

Everett Gorel
Director/ DP
South Coast Film and Video
Houston, TX 77081
www.scfilmvideo.com

Eki Halkka wrote:

>> One more thing - I always try to keep the backdrop area as small as possible, as the larger the blue screen, >>the more it will reflect blue light to the subject. I usually have some black Molton stage curtains at hand, >>that I use to cover the unnecessary portions of the backdrop.

This is the biggest single issue - the most spill is actually bounce from the screen itself.

trimming in with black as much as possible is the biggest help... and the trimming should be done AT THE PLANE OF THE SCREEN with big blacks, not just behind the actor at edge of frame

Weingartner
dp
vfx
etc
blue screen/green screen supervisor for more years than I care to remember

I’m having huge difficulty with this thread...

I’ve done a few blue/green screen jobs, that British understatement just in case you don't get it, and I've always found that the more saturated the light source the better.

Get 40% from the key colour and you should get zero from the other colours, colour separation, you know!

At a 40% level there should be zero spill and any there is should be really easy to kill because it’s such a narrow spike.

Just in case you don't think I know how to shoot for key check out my web site.

Cheers

Geoff Boyle FBKS
Cinematographer
EU Based
mobile: +44 (0)7511 506048
www.gboyle.co.uk

>> Regarding the blue and green spike tube discussion, I've used plus green or CTB on tungsten lights in the >> past, particularly when necessity dictated shooting

All good points, but keep in mind - the tubes in question have a much deeper colour than a plus green or CTB. If I may, the colour effect of such tubes would be more fair compared to an effect of a "rock’n’roll" density of a green or blue gel.

Cheers,

Pasha Patriki csc

It very much depends on the camera, and the VFX Sup. For example no one would ask for the screen to be a stop or two below key if they were shooting blue screen on a RED One camera lit with tungsten, as that was a known recipe for grainy keys. Shooting digital I always prefer to screen to be at key or a stop higher, provided that amount of light doesn't contaminate the foreground. In digital cameras the noise hides in the darker areas, so screens lit below key are usually much harder to pull.

With an experienced VFX team, there should never be any reason to light the foreground differently or with consideration for the key. People used to add magenta back lights to help separation etc, with good VFX artists this isn't needed, and actually can be to the detriment of the final composite. Light the foreground how you want it to look; you never see the green screen in the final shot.

But as I said, it comes down to the VFX Sup's personal preference and experience, and if they ask for something different (higher, lower), I always give it to them.

Jacob Dyer
Camera Operator, Post Production
Melbourne, Australia

Hi guys, long time lurker chiming in.

Agreed. The way I've been directing DPs to light screens is as per VFX Handbook. Key is irrelevant to the exposure levels of the screen, as the shadows and highlights of the foreground subjects need mattes too.

I ask that the screens be lit just below the shoulder, which depending on the camera is around 65-70IRE (bright!), so that the relevant colour channel exposes as close to true white as possible, without getting into super whites. Contamination is best dealt with by proper distance of subject to screen and limiting the amount of screen used with blacks at the screen surface. FG lighting has to be determined by background plates or by lighting design of virtual sets. Underexposed screens are way harder to deal with than spill suppression, of which there are many excellent tools and techniques.

Done this way, we get mattes that basically key themselves. Would love to hear if folks think this is wrong, as we should be suggesting revision to the manual!

James Tichenor
visual effects supervisor

Yeah OK, I'll argue with the Manual!!

At 65-70 you will not just get the key colour but some elements of non key
colour.

At 40 for key colour you will get little or no other colour.

Your key will be as good and your spill will be 2 stops less.

Can't see where that is wrong

Cheers

Geoff Boyle FBKS
Cinematographer
EU Based

Agreed, for colour difference keying you're right. But what you lose by exposing the screen so low is the opportunity for the compositor to use luminance keying, which can be a powerful tool. Rare is a matte built from a single key -- comper might start with a colour diff for the main matte and then use a lumikey for an edge and a straight fg/bg difference (assuming the camera is locked and the super remembered to shoot a clean screen pass) for some part that's not responding to the other methods -- which I guess is technically still a colour diff op but done bypassing the keyers math. The IBK in nuke can do amazing things, even without locked cameras or clean screen passes.

In the old film days, underexposed screens were a disaster because of grain -- they would basically become unkeyable. I'm not up on all the different digital cameras well enough to know about noise issues at lower exposures.

And there are a lot of great spill suppression tools and techniques to counter the hotter screens, on set and in post (without resorting to the magenta/yellow backlights).

The handbook does an excellent job of explaining the theory of a hotter screen -- in a perfect world you want to punch a hole in the screen's colour channel -- I've pulled mattes using a reorder and colour correct node. Kind of magic when it works!

James Tichenor
vfx supervisor
Vancouver

>> I ask that the screens be lit just below the shoulder, which depending on >the camera is around 65-70IRE >> (bright!)

This is the way I light green screens when I can, and the keys are super clean. It depends on how big the screen is, though, and how big the stage/room is that I'm shooting in, what colour the surroundings are, etc. Green/blue tubes in a small stage/room result in a LOT of spill.

Clearly if the screen is full body including floor then the rules are different, but I light the screen as brightly as I can for the environment if the subject isn't interacting directly with it. The rule of thumb I was once told by a guy who made a career out of shooting blue/green screens is that most keyers want to see 40 IRE of separation between blue/green and the next closest colour for optimal results. I'm sure things have improved since he told me this, but that's what I aim for.

Having said that, I've shot green screens at as low as 30 IRE in adverse situations with good results as long as the screen is even.

I think the 40 IRE exposure rule is probably left over from film, where the film colour gamut allowed for a bright blue or a saturated blue but not both at the same time. Exposing a blue screen at 18% reflectance maximized both those parameters, and 40 IRE is basically where 18% gray falls in Rec 709 video. (It's actually something like 41.7 IRE.) As the Rec 709 colour gamut does allow for bright -and- saturated blue at the same time this technique is obsolete.

I think that as long as the screen is as bright as possible and as even as possible you're 80% there. Collared tubes are very efficient as you need fewer lights to illuminate the screen: fabric screens and paint always reflect more of the spectrum than blue or green so lighting them with blue or green light results in the screen reflecting only those wavelengths. Lighting screens with white light when you don't have to is just wasting power.

The exception to this is when the pure colour bulbs are too intense for an small environment and cause a lot of spill, in which case I gel tungsten lights or HMIs. For blue screen the heaviest density of CTB that you can reasonably use helps a lot. For green screen I use Geoff Boyle's recommendation of Lee 122 "Fern Green," which works marvellously.

Naturally I'll do whatever the VFX supervisor asks me to do, but more and more I end up being the VFX supervisor on spots where the production company is doing the keying themselves. So far what I've described above keeps me working.

Art Adams | DP
San Francisco Bay Area

Jacob dyer wrote:

class="style5" >> It very much depends on the camera, and the VFX Sup. For example no one would ask for the screen to >>be a stop or two below key if they were shooting blue screen on a RED One camera lit with tungsten, as

class="style5" >> that was a known recipe for grainy keys.

Good point - most digital cameras are daylight balanced natively, so tungsten balance means upping the gain on the blue channel, which means noise. Also, most current cameras use a bayer pattern sensor, which means there's twice as many green photosites compared to blue ones (not to mention the compressed video formats, which can throw out 75% of the resolution in blue and red colour difference channels at worst). These are the main reasons why the industry has shifted from blue to green screens, AFAIK. Along with the fact that blue clothes tend to be more common than green of course...

As far as the exposure for the greenscreen goes, I prefer to light it to a neutral level - Not too bright (spill), not too dark (noise). The actual IRE value depends on the physical screen colour though. Especially as I do quite a lot of full body shots with greenscreen cyc, where the talent actually stands on the screen.

For example, Rosco's digicomp is brighter and more saturated than their regular chroma paint. It caused some spill problems, and instead of trying to change lighting to compensate, we changed the paint at our studio, to the regular Rosco chroma green. We get much less spill problems with the duller paint - and it's still plenty saturated enough.

>>> With an experienced VFX team, there should never be any reason to light the foreground differently or

>> with consideration for the key. People used to add magenta back lights to help separation etc, with good >> VFX artists this isn't needed, and actually can be to the detriment of the final composite.

That would be my advice too. Trying to fight spill with opposite colour lights will likely just cause problems.

Where too much spill becomes a problem is extracting the matte - as said, the camera (or the keying software) can't see the difference between a semi-transparent object and an object that has spill in it. Both will be of the same colour when shot and become similarly opaque or transparent in the process.

But spill isn't really that much a problem as far as colour goes, that's pretty well fixed by modern keying software. The software subtracts the backdrop colour from the foreground object - and this is where the
collared fill lights become a problem - instead of getting a neutral result, you will get foreground object with a collared fill light - just like you shot it.

There's quite a lot of stuff involved in this all, and as said, the post folks all have their own preferences. I personally have been lucky, in most cases I also do the post on my own green screen shoots, and if
someone else shoots the footage I do post with, I usually am still on the set, but serving as the VFX sup.

As said, my own preference is to light the backdrop with neutral colour, neutral levels. That doesn't make the other preferences wrong though...

Eki Halkka, DP / Jack Of All Trades, Helsinki, Finland

>> For green screen I use Geoff Boyle's recommendation of Lee 122 "Fern Green," which works marvellously.

And for blue screen Rosco Moonlight Blue which looks cyan to the eye butpure blue to a camera.

I've shot blue and greenscreen for 30+ years and it’s got easier and easier but the crazy thing is that if you apply the rules that I learned 30 years ago today’s VFX guys will be buying you drinks!

There are tests on the website of different exposures for green and blue screen complete with density readings from RGB, these came from me being paid to test every film stock combined with every screen and every lighting solution so that I could manage a Lego job that had translucent green/bluecomponents in front of a screen.

I only posted one film stock but I did test everything, hey! I have to keep a few things to myself...

Cheers

Geoff Boyle FBKS
Cinematographer
EU Based

James Tichenor wrote:

>> Underexposed screens are way harder to deal with than spill suppression, of which there are many

>> excellent tools and techniques.

I am not going to disagree with James on this, however....

Most people just want "the" answer.

The problem is that the answer is highly situational.

I'm another one who has been involved in supervising and shooting blue and green screens for film and electronic cameras for nearly thirty years. During that time, the compositing tools on the electronic side have evolved tremendously, so that the techniques that made sense for chroma key (applying light tints of complementary colours for back light, for instance) made a lot less sense once more sophisticated hardware Ultimatte versions and various software chrominance keying tools became prevalent.

I was also involved in testing and developing different fluorescent phosphor combinations for use in narrow band tubes for green and blue screen, as well as fabric choices.

Here's the objective part: As the software tools have developed, in many cases, the "hard thing to fix" which used to be spill (or colour contamination) became much easier to fix, so the tolerance for foreground colour contamination has gone up. At the same time, as keying software has become more sophisticated, the need for colour "purity" (which is related to saturation of the intended colour relative to the saturation of the other two colours) has become less critical. At the same time, the luminance and evenness of the screen has become less critical - we used to try and stay within one or two tenths of a stop over an entire screen - now nobody takes that kind of care (and for good reason)

Here's the subjective part:

As the old saying goes, "If what you have in your hand is a hammer, all your problems look like nails."

1/. Different vfx supervisors, coming from different post houses, or different types of experience as compositors, will have different opinions as to what it the "ideal" screen brightness and saturation. If you worked at a post house that had to develop proprietary software tools to pull mattes from underlit screens, you are likely to allow lower screen levels as long as they are even because you know how to deal with that more than will spill. If your experiences is with really great spill suppression tools, you might accept a higher level of contamination.

2/. The subject matter makes a huge difference! a dark -skinned bald guy in a black leather jacket offers a very different situation than a frizzy blond woman smoking a cigarette while drinking water out of a clear glass.

3/. As we revisit on CML every five years or so, the way people meter and describe screen brightness is not uniform, so when someone says, "I want the green screen half a stop under" this is not a meaningful statement on its own. As measured directly with a spot meter? Read with a gray card in the screen lighting compared to one in the foreground lighting? Read with an incident meter? (bad idea)

Now that we are largely having this discussion in the digital camera world, the discussion is a lot less messy - we can look at scopes instead of spot meters, which means we are looking at what the camera is recording, not having to extrapolate what neg densities we would be achieving on film...BUT

The combination of lighting and fabric or paint results in what you are shooting. If you have really good monochromatic fabric screens, you will probably be fine with wide spectrum lighting, but if you have really crappy fabric, narrow band illuminators will make much more of a difference. The discussion of whether you have better or worse spill from narrow band versus wide band illuminators is based on incorrect illumination levels or a misunderstanding of the physics involved.... and many people in this industry draw false conclusions from their experiences... or rather, make false generalizations based on particular situations.

Having said that, I am greatly amused by the "it works in practice, but will it work in theory" crowd. I get paid a lot of money to get shit right, so notwithstanding all the academic background and experience and theory I bring to work every day, I would much rather do a quick and dirty test and offer it to my vfx vendors and let them pick what they like the best. I rarely get that chance anymore, so I have to rely on experience and theory more, but I am, at heart an empiricist.

Forgetting a situation with a green (or blue) cyc and floor, where you must light with wide spectrum sources so that the talent is properly lit, my personal experience has been better in touchy situations with narrow band lighting on screens, EXCEPT for one particular situation:

On the movie Mr and Mrs Smith we had a lot of sets that used blue or greenscreen. For that film, Bojan Bazelli, the DP, was shooting one of my least favourite film stocks for process work - 5277, a somewhat colour-muted 320 speed stock. Pastel colour rendition usually goes along with bad crosstalk between colours.

A dull stock - never liked it....

BUT... Bojan had come up with a way of getting just the look he wanted. He was pounding the set with hard light and WAY overexposing - printing in the low forties.

This gave him the look he wanted... it also gave us a film stock exposed at a level that resulted (when we used narrow band bulbs) in some odd artifacts at black or white edges against the blue and green screens... and it turned out that when lit tungsten with no gel, the artifacts went away...
so counter to my previous experience, these tests led us to light our screens with tungsten lights.

So What is the conclusion?

Between the subject matter, the taking camera, the end product, the foreground lighting, the post house preferences, there is no "right answer" but there are some preferences that make good sense.

In general - fabrics or paint that reflects a narrow spectrum are preferable to fabrics or paints that are broader band for the simple reason that the narrower the colour band that you are using to key from, the easier to include all of it and exclude all other elements in the frame that might have some of that colour in them.

In general, the more you can mask the screen AT THE SCREEN to reduce it to the area needed for a specific shot (or at least turn off any portion that does not need to be lit) the more you can reduce specular and non-specular contamination as well as reducing veiling in the lens.

In general you are looking for an exposure that offers a good balance between high enough exposure to make good colour difference between edges and the screen and low enough exposure to minimize contamination and spill. That balance will skew up or down depending on the specifics of the situation and the preferences of the post house based on the tools they have
Off the bat, looking at rec 709, 40-50 units is where I often land, but it all depends.

If you are massively over or underexposing (e.g. setting the EI of an 800 speed camera at 1600 or at 120) it is worth spending a moment to consider the effect on the matte.

In general, if you do what the vfx person wants you to do, you will be blamed for fewer problems, even if they are asking for something that you know is sub-optimal.
I'm such an arrogant fuck that I will pretty forcefully try to persuade an inexperienced VFX person to let me shoot what I think will be optimal, but your mileage probably SHOULD vary... my manner may not be appropriate for all situations.

Mark Weingartner
la
dp
vfx
etc

In other words, every screen, setup, shot, foreground, background, stage, lighting package, camera and situation demands a specific answer. Couldn't agree more.

Unfortunately in prep the DP will always ask, what screen colour and where do you want it exposed, and me answering "it depends" just doesn't fly. So I'll say, generally I like 'em just shy of 70 but we should talk about it in each case. Inevitably they'll look at me like I'm insane and I'll spend the rest of the show trying to convince them that

Great post Mark, should be printed and handed out to all the VFX Supervisors currently studying in school!

James Tichenor
vfx supervisor
Vancouver

Something I'd add to Mark's awesome post, hopefully without stepping on it, is that especially for blond hairy heads, regardless of what flavour of ProRes or other compressed file you may be shooting for the rest of the show, consider shooting the screens uncompressed, or RAW, or 4K for 2k. Prores 4444 is awfully close, maybe close enough not to make much difference, but generally speaking, compression nibbles at edges.

Another thing that's been a VFX staple for years is to shoot a tight shutter, and add digital motion blur back in the comp. Spill on things like shiny black shoes which get motion-blurred can require tight rotoscoping to hole-fill rather than just a garbage matte. Half or quarter will do it.

And please don't stick green cards on blank monitors, unless being seen through talent's hair. Just small tracking dots if it's black on black. We only need green/blue to matte irregular foreground objects. And with hard static edges it's much faster to just draw and track in a matte, with no spill worries.

Tim Sassoon
SFD
Venice, CA

>> In general you are looking for an exposure that offers a good balance between high enough exposure to

>>> make good colour difference between edges and the screen and low enough exposure to minimize

>> contamination and spill.

I find that having a component 'scope of some sort, whether an RGB parade or overlay waveform monitor, or an RGB histogram (as on the RED), is very helpful: you can see directly what the colour component separation/isolation is, as well as the individual RGB levels, so you can dial in the light with good separation and component gain, yet without overdriving the colour channel you're keying with (which is easy to do on bluescreen if you're only monitoring the luma level).

(Mind you, if you're shooting film, I haven't yet quite figured out where to plug in the waveform monitor. But for video / digital cine, I find the 'scopes invaluable.)

Adam Wilt
technical services: consulting / coding / camerawork
tech writer: provideocoalition.com/awilt/
Mountain View CA USA

To answer some questions the screen I'm working with is 120' wide and 13' back from a raised set. It's a night scene so we opted for blue so spill was less offensive if any was left in the final composite. In the past I used to expose the green or blue screen 1 stop under key for film and at 50% on the waveform for video. Lately I've been using the waveform to check for evenness and the vector scope to set maximum legal chroma when using blue or green spike tubes. This makes the blue screen a bit hot on the waveform. Is this a bad practice?

Regards,

D. Gregor Hagey, CSC
DOP
Toronto

I agree with Geoff.

As more of a "post guy" my understanding of the way Ultimatte works (not sure about other algorithms or methods for other keyers):

The way stuff gets keyed is by comparing the RATIOS of green to red and blue in the foreground and the background (for a greenscreen). In ANY footage of a green background, the pixels that are "green" will also have SOME amount of red and blue. When you have levels as high as 65-70, the only way that that brightness happens is to have a LOT or red and blue in the "green" colour. The same thing happens at the other end of the luminance "spectrum." As you get darker and darker, the RATIO again begins to get smaller and smaller, because even if you ONLY have green channel level (with no blue or red) in the "green" colour, the RATIO has become close again because there's not much green level to begin with.

This is why, when lighting green screen you don't just want "a waveform monitor" to judge exposure, but an "RGB waveform monitor." The RGB waveform monitor shows you all three colour channels and you can see the actual ratio between them. Starting at black, all three colour channels have the same level. As the luminance raises, the green (or whatever colour the background is) will be stronger than the other two colour channels, until, at a certain point (probably around 40 or 50IRE) the other two channels will start to rise in comparison with the colour channel for your background until the background goes completely white, at which point the levels of RG and B are again all equal. The goal is to light/expose to get the highest ratio of the green compared to red and blue channels (when shooting greenscreen) or blue compared to red and green (when shooting blue screen) The exact "best" exposure might depend on the specific hue of the background.

I had heard the "conventional wisdom" of shooting greenscreen with a magenta backlight or bluescreen with a straw filtered backlight (because magenta is the opposite of green and yellow is the opposite of blue), supposedly to nullify any spill/bounce from the background, but I think the problem is that this actually makes the key look hard and fake on those lit edges. I have to say that that is second hand info from the VFX guys...

Steve Hullfish
contributor: www.provideocoalition.com
author: "The Art and Technique of Digital Color Correction"

Here's a bit more info. The digi bluescreen is 120' wide and 13' back from a raised set. The lights are well flagged (all blue spike kinos at 1 bubble low). I'm recording to gemini from Alexa at 800ISO.

In the past I would expose a blue screen at key or 1 stop under key for film and between 45-55 IRE for video. Currently I use the waveform to judge evenness, but use the vectorscope to set the exposure on the blue screen in order to get saturation as close to the blue box as I can. This makes it look a bit hot on the waveform, but to the eye on the monitor it's very, very blue. It sounds like no one uses the vectorscope and maybe I should forget chrominance and just stick to luminance on the waveform?

Regards,

D. Gregor Hagey, CSC
DOP
Toronto

James Tichenor wrote:

>
>> Agreed, for colour difference keying you're right. But what you lose by exposing the screen so low is the

>> opportunity for the compositor to use luminance keying, which can be a powerful tool.

If you light the screen very bright, you can use luma to key dark parts of the foreground subject, yes. But what if your talent has a blond backlit hair, white shirt and holds a flashlight?

- if the screen is lit very bright, that can help luma/difference keying dark foreground subjects (but not bright).
- if the screen is lit very dark, that can help luma/difference keying light foreground subjects (but not dark).
- if the screen is lit to the center, that can help luma/difference keying both dark and bright foreground subjects (but not middle exposure).

>> I've pulled mattes using a reorder and colour correct node. Kind of magic when it works!

Me too - I’ve also done spill correction by combining channels manually. As well as working in YUV or other more exotic colour spaces for keying. And lots of other similar stuff...

Slightly OT: In fact, I did stuff like that so often that I ended up writing my own little keyer for After Effects, that does some of the tricks I used when the other keyers didn't work out of the box... Sadly,
Adobe dropped support for Pixel Bender language in CS6, which made all my plugins die. Oh well... they still work for CS4-CS5.5. In case someone wants to try it out, here's a zip with it (along some other plugins I wrote). Your mileage will vary, sometimes it may actually work better than other keyers at least on some parts, more often not.

http://eki.pp.fi/temp/Eki/HalsuPlugs_20130128.zip

Eki Halkka, DP / Jack Of All Trades, Helsinki, Finland

An RGB Parade waveform monitor is a better choice than a vector scope for this. Another tool that might be even more handy, though I have never tried it, is Tektronix's Diamond scope which shows red and blue channels on one side of the monitor and green on the other.

The box on the vector scope means less than you think it would, though it would certainly help dial in the correct hue.

Disclaimer: I sometimes consult for Tektronix. But many manufacturers make RGB Parade waveform monitors and any of them would work fine for this purpose.

Steve Hullfish
contributor: www.provideocoalition.com

>> I had heard the "conventional wisdom" of shooting greenscreen with a magenta backlight or bluescreen >> with a straw filtered backlight (because magenta is the opposite of green and yellow is the opposite of

>> blue)

As best I can tell this is left over from the days of chromakey. I've never seen this done successfully, in a way that looks natural.

Art Adams | DP
San Francisco Bay Area

>> The biggest mistake I see is over lit screens using blue or green tubes. With a properly lit screen, I don't

>> understand your VFX supervisor's statement.

The official answer is: it depends .

Using blue or green Kino tubes, normal metering is useless because those tubes have virtually no luminance output, so the meter way underreports what is there. A *real* waveform monitor will help get proper exposure.

With tungsten instruments, I like blue at stop and green half a stop below, but it also depends to some extent on the quality of the screen or paint. Most screens and paint out there are total crap. The stuff from Composite
Components in Pasadena works well for me, has for decades.

Bob Kertesz
BlueScreen LLC
Hollywood, California

"As best I can tell this is left over from the days of chromakey. I've never seen this done successfully, in a way that looks natural."

Regardless of its derivation, I've used the technique successfully, straw with bluescreen, when compositing a person against a warm, sunny background plate. For my taste, I think it can actually look more natural than white light. As far as using magenta with greenscreen, regardless of how well it may key, magenta backlight just doesn't generally exist in too many places on this planet. So I agree with you there.

Everett Gorel
Director/ DP
South Coast Film and Video
Houston, TX
www.scfilmvideo.com

Everett Gorel wrote:

>> I've used the technique successfully, straw with bluescreen, when compositing a person against a warm,

>> sunny background plate.

Makes sense if you have a warm sunny background plate.

In the modern world, the backlight colour, quality and intensity should be driven by the plate or creative choices.

I recommend that DPs turn off the greenscreen and look critically at the FG lighting periodically as they work.

Weingartner
la
dp
vfx
etc

>> Regardless of its derivation, I've used the technique successfully, straw with bluescreen, when compositing >> a person against a warm, sunny background plate.

That's not using straw to prevent contamination, that's simply matching a background plate. I'm sure it helps, but you're not really using it just to eliminate spill.

I'm not a fan of straw in HD except under certain circumstances. It can tip over into green too easily.

Art Adams | DP
San Francisco Bay Area

Geoff Boyle wrote:

>> At 65-70 you will not just get the key colour but some elements of non key colour.
>
>> At 40 for key colour you will get little or no other colour. Your key will be as good and your spill will be 2

>> stops less.

I'm shooting green screen with green spike Kinos as a bkg source today and I have screen level on the WFM at 52 ire mas o menos. The trace is flat and thin. Subject key side is about 62 ire .

On rgb parade green is about 80 ire , red about 10-15 and blue just above 20 for those of you into ratios. Key is virtually one click with Fusion . Oh yeah, 2 canon c 300's > Pix 240 recorders recording pro res 145 Mbs.

I do agree with Mr. Weingartner about the situationality of green screen. This could be a disaster with different spatial relationships between screen, subject and lighting involved, but in this case the green screen/ key concerns are below minimal.

Mark smith
DP NYC

If the trace is flat and thin, that means you're in the "wrong" mode (at least for judging saturation). Though the ratios on the RGB sound about right.

If you have saturation or chrominance and the waveform is flat, it's because you're in LOW-PASS mode. Flip over to Flat Pass and the background should display a nice thick excursion... unless you're shooting a white luma key instead of a green screen.

A good way to tell when you have a white balanced camera is if you have a flat, thin trace and you're in FLAT PASS mode. If you're in Low-Pass, it's impossible to tell the difference, but it's easier to see your luminance. In FLAT PASS mode, a thin line is a sign of NO saturation.

FLAT PASS means that the signal isn't filtered for all of the high-frequency chroma information. Low-Pass lets the lower frequency luma information "through" and filters out the high-frequency stuff.

Trivia: the "thickness" of the trace is called the excursion.

Steve Hullfish

Hi There,

Having shot and worked on green screen and blue screen footage for over a decade, I have to say that this thread is one of the most valuable I ever encountered on CML.

Thanks for all the input.

I guess I need to compile some meaningful tutorial from it (and Geoff's test results) for my team and trainees.
There is a lot to learn and it feels some hundred years of experience at hand.

Mit freundlichen Grüßen,
Best regards,

Axel Mertes

>> If you have saturation or chrominance and the waveform is flat, it's because you're in LOW-PASS mode.

Also, if the regular waveform reads 52 IRE but in RGB parade green reads 80, then the first reading is clearly low-pass/luma only. That luma waveform is derived by taking 71% of the brightness info from the green channel and adding it to 21% of the red channel brightness and ~8% of the blue channel brightness.

52 IRE just tells you how bright the screen is overall using a weighted average of the brightness in the three colour channels but it doesn't tell you anything else about the colour. Parade tells you the weights of the actual colour channels. In this case, 52% is how bright the screen is if you ignore chroma, but 80 IRE green compared to red and blue at 10-20 should result in a very nice key.

Art Adams | DP
San Francisco Bay Area

>> That's not using straw to prevent contamination, that's simply matching a background plate. I'm sure it

>> helps, but you're not really using it just to eliminate spill.

No, I don't use it "strictly" to prevent contamination. That's why I made the distinction about using it when it makes sense (matching certain daytime exteriors), and conversely, NOT being able to generally use magenta with greenscreen (even though it's a technical complement) unless you're compositing against neon or some other planet:-) Again, when it makes sense.

Regarding the straw tipping over into green, though I've never personally experienced it, I appreciate the warning. Thanks.

Everett Gorel
Director/ DP
South Coast Film and Video
Houston, Texas

>> Regarding the straw tipping over into green, though I've never personally experienced it, I appreciate the >> warning. Thanks.

It can be a little hard to see until you become acclimatized to it. For a long time it really only looked really yellow to me, but an obnoxious yellow, and that's when I realized that there's a very fine line between bright yellow and green. It's easy to cross it with heavy CTOs and straws and saturated colours like fire.

I most recently saw this look in The Hobbit, whenever they'd rake tungsten "sunlight" through a set and it was striking someone from 3/4 back. It wasn't a warm orange/yellow, it tipped over into yellow with a green tinge that looked really unnatural to me.

When I do something similar I'll often add 1/8 minus green to get rid of that slight imbalance.

I can see it on my monitor in the trailer:

https://www.youtube.com/watch?v=nOGsB9dORBg

...in the yellow sky at 1:30. There's also a very quick shot at 1:52 where the yellow highlights are on the edge of green but not too bad.

In a way it's like looking at a sunset and waiting for the green flash: at some point the yellow just goes over the top and turns green.

Not a great example because this looks fine, but...

http://www.scmp.com/sites/default/files/styles/980w/public/2012/12/31/the_hobbit_an_unexpected_journey_

movies_the_hobb_32983803.jpg

One or two more IRE in the green channel and that would look really odd if it were a video image.

It's interesting to look at that image in Photoshop... there's no point in the image where the green value of a pixel exceeds red. Someone was very, very careful.

Art Adams | DP
San Francisco Bay Area

"in the yellow sky at 1:30. There's also a very quick shot at 1:52 where the yellow highlights are on the edge of green but not too bad."

https://www.youtube.com/watch?v=nOGsB9dORBg

Then you must LOVE his edge light at :47. Maybe we just chalk that one up to motivation by the lens flare.

Thanks, as always.

Everett Gorel
Director/ DP
South Coast Film and Video
Houston, Texas

>> Then you must LOVE his edge light at :47. Maybe we just chalk that one up to motivation by the lens flare

The guy on the horse? That doesn't bother me too much. If the scratch is greenish then it's pretty subtle and you could write it off to sun hitting grass and leaves outside the frame and reflecting green light into the shade. (The best example I saw of that was the house exterior in the first PGA camera shootout. )

The flare is greenish but that doesn't bother me either because it's just a flare.

I'm more bothered by shots where it's clear that it's supposed to be warm sunlight but it's so warm that the yellow becomes more green than red. That really annoys me.

Art Adams | DP
San Francisco Bay Area

Steve Hullfish wrote:

>> If the trace is flat and thin, that means you're in the "wrong" mode (at least for judging saturation). >>Though the ratios on the RGB sound about right.

Oh sure. But you know when looking at green screen luminance the thickness of the trace tells what you need to know about difference in illumination in the vertical dimension just as the shape of the waveform horizontally describes luminance across the frame from left to right.

No matter. The keys were effortless.

Mark Smith
DP
Ex-NYC as of now.