Green screening action footage with DVX100B

A

Axem

Member
Hey everyone, this is my first post here. This is an awesome forum :) I hope to learn a lot here.

I'm currently trying to shoot green screen with my 100B and am not having very good success. The types of shots I'm shooting are things like fight choreography, free running, flips, martial arts, etc, so its all very fast motion. The problem I'm running into is once I get it into FCP keying out the green becomes very difficult. We shot on 24PA so fast movement becomes blurry and messes up the green screening. What would you guys suggest? Should I be filming on 60i for stuff like this? This is my first time using a green screen so I know I still have a lot of learning to do, but most of the threads or tutorials I've found usually just focus on stationary subjects talking or something related, not fast movements and action.
 
60i for green screening is terrible :)

put it in 24pa and throw the shutter speed up high enough so that motion doesn't blur
 
the problem is separating foreground from background in the 4:1:1 color space (DV).
eliminating the background, then replacing it with another source, while keeping the foreground 'present' (clean) is a tall order.
*there's this nice plug-in in FCP - provided by a third party (sorry, not in front of my machine right now) it's in the 'key' list there. it works great, and is about the best you're gonna get at doing the job.

lighting is important to pulling it off better (the more you do to perfect the process, the better results you will obtain).
making sure the background is lit evenly is a start.
if the background can be further away (to be out of focus), that's good.
light your foreground so that it is nice and even, with proper key and fill lights, but be careful about spilling over shadows onto the background.
separate the foreground characters from the background by putting a little ring light (highlight, or rim light) with a straw or wheat gel on it.
*the idea being to DEFINE the edge between what you want to key from the background.

shoot 30i, 4:3 aspect ratio (you get the highest level of detail there).
if you want wide screen later, just make overlays in FCP at the final cut (you can make guide lines on your viewfinder for framing).

since what you're capturing is faster action, you might want to play with finding the right shutter speed (higher the frame rate, add more light).

it's lots of little nudges to get the thing right.
not a finite science (working with what you've got).
if it's really important, you should consider investing in higher grade gear to accommodate.
look at the ULTIMATTE (web page); http://www.ultimatte.com/UltimatteMain/HOME/HOME.html
*center of the known universe of keying. HD/SD/etc.

here's another good resource; http://www.seanet.com/~bradford/blue_green_screen_visual_effects_1.html

you'll get a lot of good info there- in your journey.
report back with results- and good luck.

*oh yeah; it's best if you can really control the light in the shooting space.
it's really all about the light in the end. you'll see.
 
Awesome, thanks SO much for the help guys :D. Really appreciate it. Will try it out and report back.
 
NoNoob said:
shoot 30i, 4:3 aspect ratio (you get the highest level of detail there).
Click to expand...

No, you won't. The highest level of detail you can get from the DVX is with progressive in "THIN" mode. Interlaced will never be more than about 77% of the vertical resolution of progressive.

As for whether or not 4:3 or squeeze would be better, there are arguments to be made either way.

You always want to shoot progressive if you want to do greenscreen because you'll do away with the interlace artifacts inherent with interlaced. Those will mess with your key as much as or more than the dithering effect you're seeing in your motions.
 
Yeah, you can try magenta- (experimentation is beneficial).

But (in my experience) magenta is a bit out of range for the separation (in consideration of the chromatic frequency of key green- if it's truly in that value).
*if you have a scope handy you can see their traces easy enough.

The wheat/straw gel (golden color) is at the diametric opposite side of the normal key green color.

Fringing found in just straight lit scenes (especially around hair) is limited by the value differential using the 'gold' color.
Magenta leaves it's color quotient in that edge. It gets hard to dial out (normalize).

Trust me, we tried it - and lots of others.
Straw (AP2240).
I don't have my equiv. book handy right this second- but it's easy to find for Rosco and others.

The principal of the rim light is to 'pull' the difference between foreground and background.
Think of it as a 'knife' to help the key cut (making the software work easier).

Show us some clips!
 
NoNoob said:
well, i guess i been told- eh?

just try it my way before you go doing something stupid (and waste your time).
Click to expand...

Dude . . . are you seriously saying that shooting progressive instead of interlaced for greenscreen is "doing something stupid"?

:huh:
 
When did "dude" become baiting? It's just . . . something people say.

The chips in the DVX are native progressive chips, and their potential native resolution is significantly higher than 525 lines (as Andromeda users will attest). It's downprocessed to DV, and then further when it's made into interlaced.

The Kell factor of interlaced is inherently lower than progressive because interlaced video must be blurred in order to eliminate flickering. This happens when a detail is small enough where it appears on only one scan line, so it's on only one interlaced field and not on the next. It sequentially appears and disappears, causing a flicker. To avoid this, a "low pass filter" is implemented and the detail is effectively blured onto two lines, thus lowering the resolution and detail.

This is not a problem with progressive video, because the entire picture is drawn every time and there is no flicker. No blurring is necessary.

Thus, interlaced video is 25-30% lower-resolution than progressive video.

This is why the Vertical Detail Frequency exists. Progressive video shot can flicker on interlaced screens because the detail is too high. Switching to THICK solves the problem by lowering the detail to that of interlaced video.

Progressive footage at the THIN setting is the highest detail the camera can achieve.

That, coupled with the interlace artifacting of interlaced, makes progressive the obvious choice for keying.
 
Hey all, just finished capturing everything and playing with it in FCP.

Results so far:

Shooting on 24PA with 1/250 shutter + thin vdetail worked GREAT! Soooo much better than the day before when it was just on 24PA. The only problem with shooting with a high shutter was that the fluorescents in the room (we're shooting in a gymnastics facility) gave a 'pulsing' effect on the footage and continually changed the green to different shades. This is a minor inconvenience at best because I just need to throw on two color key filters and get both shades of green.

We will be shooting again this weekend and I will try out NoNoob's suggestions. I want to try every method and setting ppl think will work so I can see what works best (I'll edit a quick video after to compare the results).

Thanks again to everyone for their help. I do have one last question though: what software do you guys think is best for keying? So far FCP has been adequate, but I've heard people say that After Effects is better, whats your guys' opinion?


Edit: Here's a youtube video I made last year, mostly shot on DVX. This is the kind of stuff I'm trying to film with this greenscreen project. We're planning on using the CartoonR filter to create an online animated series incorporating things like this. It's nice because the cartoon filter masks imperfections in the greenscreen shots (such as the halo around subjects, etc) so we don't need to be perfectionists with it.

http://www.youtube.com/watch?v=02aEE8M8RVA
 
Last edited:
Your video example- GREAT STUFF! WOW! That's incredible footage! Amazing feats.
*A compilation of many segments. All shot by you or your team?? Awesome!

Your recent better results might have come from the higher shutter speed, not just 24PA. Do your tests and let us know.
It'll be interesting to hear.

KILL THE FLORESCENT LIGHTS! they pulse at 60 cycles- bad for business.

The higher the shutter speed, the more LIGHT you need (bigger bulbs!)
The LIGHT makes the most difference in the final analysis.

30i lends itself to a more manageable result for the type of output normally encountered in our realm- prosumer/industrial video.
Mainly because of what you're going to do to it in the NLE.
It's just a VW Bug, not a Ferrari. Can drive on the same race track, but not exactly the 'same' results.

AE and FCP have basic color keying effects.
AE has a little more horsepower- but some don't care for it's output (downstream).
[try to use the numeric scheme- so you can make a record of the adjustments for reference.
The GUI's are okay for getting in the ball park, but the numeric will give you lasting pleasure.]
I like AE because you can drop a basic formulae on the next footage for a head start.

There are more robust keying tools out there.
Check that second URL I gave you for links to those.
It's a short list- but very beneficial to your quest.

Things to avoid:

• clothing with a reflective material (in the thread or on the surface- like inks or imprints)
• anything near the frequency of the green (it's supposed to be a non occurring in nature color- but anything in the neighborhood can be a ***** to fix)
• WHITE- no white. *Midtones in gray or brown are best.

Since you're trying to capture full body movement it's best if you light using the 45°/90° lighting scheme.
Two light sources left & right at 45°, your camera at 90°- action moves before the lens.
The ring light (rim light) at nearly 90° to the talent shooting across their shoulders at each side- it's tough in full body, AND with motion... but it can be done.

If your green backdrop is further back from the action- it's important to light it as evenly as possible- top to bottom and left to right. Hand held meter helps dial it in.
Focus on a line chart in the path the talent will use, use widest ƒ stop possible.
Determine your depth of field at that specific focal plane, mark the floor or pad with tape to help them keep in the 'zone' (if possible- this stuff looks pretty free form).
Zooming in half way is better (line chart helps you find the crucial focal point).

If it's possible- bring your rig into the shooting space (the gym) so you can have realtime experiments. Laptop is okay fine for this.
You can dial in the lighting and see where the 'holes' are- on the spot.
It'll also help the talent get a feel for the process, and 'tune' their performance for best results.
Good luck!!
 
Last edited:
NoNoob said:
30i lends itself to a more manageable result for the type of output normally encountered in our realm- prosumer/industrial video.
Mainly because of what you're going to do to it in the NLE.
Click to expand...

How is this so?

Sure, it's 29.97 and the NTSC broadcast standard is 29.97.

But the overwhelming majority of prosumer/industrial work isn't going out for broadcast, and even if it were, the NLEs are perfectly capable of handling 30p or 24p and outputting them at 59.94i for broadcast.

Most people are going to put their work on the Web or on DVD, both of which can take 24p or 30p material with no need for interlacing.

What specific advantage does shooting interlaced give you?
 
NoNoob said:
the chips (3x) in the camera are 4:3, CCD, analog NTSC video (RS170A), with about 525 lines of resolution, +/-.

anything done to that signal, downstream of the taking aspect, is an affectation.
after the fact processing.
Click to expand...

You appear here to be saying that the CCDs are interlaced and that the progressive footage is created out of interlaced by processing, similar to what a "Frame" mode in other cameras would do. In "Frame" modes, you lose detail compared to the interlaced footage.

But that's not how the DVX works. It's not a "Frame" mode. It's not an affectation; it's true, progressive footage from progressive chips. And in THIN mode, it simply is more detailed than the interlaced footage. That is verifiably -- and verifiedly -- true.
 
NoNoob said:
Like I said, I'm not worthy of pontificating on the subject (I only have a couple thousands hours experience doing green screen).
Click to expand...

Why the repeated sarcasm?

Look, apparently, we somehow got on a bad footing here. If I contributed to it, then my apologies; I didn't mean to. Let's start again. :beer:

I haven't argued with anything you've said about the process of setting up a greenscreen shoot, or the ultimate process of it. I'm not at all denying your credentials, and I think your input has been tremendous.

I'm simply making a specific point about a specific camera which (after thousands of hours of working with it and its footage) happens to be fact. The highest level of detail is achieved by using progressive/THIN.

What you say may well be true for other cameras; I won't argue that. But for the DVX specifically, it isn't.

In any case, I've explained specifically why I say what I say. If there's a specific advantage to shooting interlaced, which overcome what I've pointed out as the disadvantages, why not just explain it detail instead of just saying shooting progressive for greenscreen is "stupid"? If it's helpful, it's helpful. Everyone will benefit.

What, specifically, for your purposes -- shooting, editing, FX work, final output -- makes shooting interlaced --with the DVX -- the better choice?
 
i've deleted my contributions to the thread that veered off the path to resolution for the question being asked in the original instance.
i've also modified contributions i made to keep on task.

thanks for the consideration.
p.
 
Last edited:
David Jimerson said:
The chips in the DVX are native progressive chips, and their potential native resolution is significantly higher than 525 lines (as Andromeda users will attest). It's downprocessed to DV, and then further when it's made into interlaced.
Click to expand...

wait, so does this mean if you used Firestore or some other SSD device you'd be able to get higher resolution than DV? or is it possible to record with the DVX to a codec other than DV?
 
There is only one way to bypass the DV encoding.
It is(was) a product called ANDROMEDA. Which is a $2,000 mod this guy was doing to DVXs which grabbed the output right at the CCD and sent it out a new output port.
The resolution is WAY higher... 1540x984 or maybe even higher. And it looks AMAZING.
But it was expensive and a little complicated.
I'm not exactly sure why people would do this as opposed to just getting an HVX200...
Maybe it is a matter of a different look or something.
 
You must log in or register to reply here.
Back
Top