cheapest camera for perfect green screen work

[maxotics]

A lot of credit goes to Andy who really drives the conversation deeply.  This forum would be a pretty dull place without him.  I mean that sincerely.

[dishe]

I can't see the video (due to copyright-laws in my country), but your comment makes no sense. Andy surely didn't perform the keying in 480p. 

 

You misunderstand- The video posted above is presented here at 480p maximum resolution. Aside from that, the video moves very fast and is chock full of composited effects and CG. This is hardly an example of the ability to get a clean key out of a 4:2:0 image- a bad key out of that 5D he used would be pretty much indistinguishable in this sort of situation. As I said before, if you told me that image came out of an iPhone4, I doubt the end result would look much different.

...and that's sort of what we're saying all along. A 4:2:0 camera may be harder to pull a good key than a 4:4:4 or 4:2:2 one, however you don't always need the cleanest key in every project. For a video like Andy's example, it works just fine. But "fine" isn't "perfect". Fine is adequate, fine works. Fine isn't best... its just fine.

 

 

Your explanation remains wrong. If a 1080 clip has visible artifacts, they get worse when you downsample the video. Noise is more prominent, jagged edges are magnified, every other artifacts become more visible. 

Haha, you say that with such conviction! Unfortunately, I'm not sure what universe you are referencing with this data. Anyone with experience in post production (or heck, even basic photoshop!) knows that sampling down the image REDUCES noise and jagged edges. One of the advantages of shooting higher resolution than your target delivery format (4k down to 1080p, for example), aside from being able to reframe your shot, you can reduce the image size and remove noise and artifacts. Unless you are using poor bi-cubic resampling, jagged edges will smooth out and become more defined. 

I'm going to give you the benefit of the doubt and assume that perhaps we aren't talking about the same thing- so I'll illustrate visually to help you out. 
Here's that picture from the Wikipedia link above discussing color sampling:

Here is the 4:2:2 and 4:4:4 version side by side, showing only the color resolution:

I cropped out a portion in the middle to make sure the browser won't resize the image. Quite a difference side by side here. Clearly the 4:2:2 version is going to create a messier key, with blocky edges and artifacts compared to a 4:4:4 picture. 

But, what happens when we shrink the image down to 30% of the original size?


See how the blocky edges and artifacts are virtually non-existent now? A key pulled on an image shrunken down like this is going to have much more defined edges than the full 100% sized one will!

In fact, let's compare it to the original 4:4:4 image resized to 30%:

The one on the left is the 4:2:2 image, the right a 4:4:4 one. Yes, the one on the right is still a tad sharper than the one on the left, but the difference when pulling a key is minuscule compared to the full sized one above. Jagged edges are gone, noise reduced and shapes are more consistent.
Also something to take into consideration is the fact that Photoshop's ability to sub-pixel resample (which is what is used to shrink that image down) is capable of a cleaner resample vs what the camera samples off the sensor. So, an image actually shot at that size on a 4:4:4 sample would probably be slightly less sharp, and look more-or-less like the image on the left. 

Therefore, I stand behind my statement (and the statement of others on other forums- I didn't make up this theory), that a downsampled 4:2:0 or 4:2:2 image can be virtually indistinguishable from a native untouched 4:4:4 of the same size when it comes to grading and pulling a key. 

 

 

Also true for the new 'issue' with the BMPCC: You see almost nothing @1080, but once you downscale the image, the grids pop out. 

That's something else entirely. You're talking about frequency patterns, I think. 

[andy lee]

Youtube video was posted by the RECORD LABEL as a promo item ...its is nothing to do with me that its is 480

I shot and delivered a 1080p 25fps video to them

as someone who shoots expensive videos for record labels (that video had a buget of £20,000 UK Pounds and I get paid to do it , I dont do it for fun , I am saying you can shoot prefectly usable greenscreen results with a 4 2 0 camera , , MTV are happy I dont remember them refusing to broadcast my work or any other TV stations across the world that play my videos on heavy rotation on their music channels 24/7.

Too many pixel peepers on here that read spec sheets but dont deliver product .

 

This topic started with a newbie to green screen work wanting a cheap camera not a RAW /prores data low nighmare

I have shot more Blue and Greesceen commercially than alot of people on this forum and in real world , get the job done and deliver quality product to the TV station/record label time after time .

So my advice to him was a Panasonic or Canon camera will do a prefectly acceptable job ...and you can get paid if you know what you are doing...

[andy lee]

A lot of credit goes to Andy who really drives the conversation deeply.  This forum would be a pretty dull place without him.  I mean that sincerely.

I like to do things simply and easliy and not over egg the job , yes I could shoot every job in ML Canon RAW  or with a pro res  camera etc but most jobs dont need it

so I use Pansonic and Canon cameras mainly now that are 4 2 0 it works for me!! time = money

[Axel]

Haha, you say that with such conviction! Unfortunately, I'm not sure what universe you are referencing with this data.

 

 

I reckon you are pissed a bit. Accept my excuse and let me repeat where we agree: Recording in higher resolution is good. Let's set aside if this is because of anti-aliasing through intelligent interpolation by the software you scale with or because the average of 4:2:0 equals true 4:4:4 when downsampled. Life is too short.

[dishe]

Youtube video was posted by the RECORD LABEL as a promo item ...its is nothing to do with me that its is 480

I shot and delivered a 1080p 25fps video to them

 

Figured as much. But you tried to use it as an example of how great a key you pulled from a 5D. I didn't say the fact that it was 480p is your fault! I'm just saying that link isn't a good example because it was lower resolution, not to mention the style of video doesn't require it to be as perfect of a key to be realistic.
 

as someone who shoots expensive videos for record labels (that video had a buget of £20,000 UK Pounds and I get paid to do it , I dont do it for fun , I am saying you can shoot prefectly usable greenscreen results with a 4 2 0 camera , , MTV are happy I dont remember them refusing to broadcast my work or any other TV stations across the world that play my videos on heavy rotation on their music channels 24/7.

Too many pixel peepers on here that read spec sheets but dont deliver product .

LOL- You crack me up, Andy.

Between you and me, bragging about one's experience to gain notoriety on the internet often does the opposite of its intended function. With all due respect, I may be new to this forum but you have no idea who I am or what I've done. I'll say this much: Don't be so sure you are the only person in this thread who's work has appeared on television. 

 

This topic started with a newbie to green screen work wanting a cheap camera not a RAW /prores data low nighmare

I have shot more Blue and Greesceen commercially than alot of people on this forum and in real world , get the job done and deliver quality product to the TV station/record label time after time .

So my advice to him was a Panasonic or Canon camera will do a prefectly acceptable job ...and you can get paid if you know what you are doing...

Someone was asking what the cheapest camera for "perfect" green screen was. You made a good recommendation for an all-around great camera on a budget. But you keep acting like the question was aimed only at you, and if anyone else voices their opinion it must somehow be a personal battle against you and your reputation! 

You made a great point- you don't always need 4:4:4 to pull a good key. We all agreed with you- heck, even I did! I've pulled plenty of 4:2:0 keys that were good enough for what I was doing. But would a different camera have done a better job? Yes, I still think it would. And if I were buying a camera exclusively for green screen work, I'd look for at least something that records 4:2:2, like an HMC-150. Being that the OP specified his budget is around $4k, there are plenty of cameras that shoot a lot more than 4:2:0 and DON'T require a RAW workflow. Many of them shoot AVCHD or Prores natively. No reason to compromise. 

He asked for ideas. We are giving him ideas. Yours is just as valid as everyone elses, IMO, so let's all just get along K? ;)

[dishe]

I reckon you are pissed a bit. Accept my excuse and let me repeat where we agree: Recording in higher resolution is good. Let's set aside if this is because of anti-aliasing through intelligent interpolation by the software you scale with or because the average of 4:2:0 equals true 4:4:4 when downsampled. Life is too short.

Fair enough! 

 

Also, to be fair, I didn't mean to say that mathematically the average of 4:2:0 exactly equals true 4:4:4. There's no way to accurately calculate that, really. All I'm saying is that the jagged doubled-up-pixels you get on edges with 4:2:0 aren't there anymore once you downsample. Since that initial loss of resolution/edge definition is the reason shooting 4:2:0 is detrimental to pulling a key, I'm saying that it really won't matter that much what color sampling method you shot it in once you are scaling it down anyway, in this reference at least.

 

I hope we can at least agree on that!

[Axel]

Absolutely, and I hate wisenheimers as well, must be my shadow.

 

The poor threadstarter has given up long ago.

 

That was my first experience with greenscreen: I read a book about Final Cut 2 in 2001 and chroma keying was mentioned. It looked easy. I had a bright green folder on my desk which I taped on the door. I put my VX1000 (DV PAL) on the tripod. I painted red color over my wrist, focussed on my hand and made a strangling gesture. Then I filmed myself, mimicking being strangled. Of course I had 'the hand of horror' in mind. Compositing was a matter of trial and error. I had to deinterlace before I scaled the hand, I had to get rid of the arm, I had to add the cut area with the bone (Photoshop) and animate it. It was far from perfect, but it was big fun. It was diffuse daylight, and though my hand was close to the folder, I only had a few frames with spill. I simply cut them out, that looked even more creepy (funny, actually).

 

I recommend a playful, respectless start. If you haven't done it and only read about all things that let it go wrong, you'll be discouraged.

[mtheory]

Andy, your 4:2:0 camera worked well because you used it on a commercial production with a perfectly-lit stage and crew, which the OP doesn't have budget for. A better codec is about the only thing he has to even out the equation. Don't push the kid into fire, man. 

[andy lee]

lighting is the most important part of the job - what ever camera you use

 

and then hire the best 3d artist you can as all that green has to be replaced on every cut you make - and in perspective too.

so the most amazing pull off your 4.4.4 camera is no good if you are replacing green with junk -

so we spent more time 3d modeling

designing the virtual sets -

lighting the virtual sets,

pre grading the virtual sets to fit the live action ,

texturing the virtual sets etc etc etc

layering and layering and layering rendering rendering rendering get the look

 

more time doing all of that than doing green pulls off my 4 2 0 camera as thats was the easy bit!

 

green replacement takes time - lots of time , its not always fun waiting for massive renders

[Axel]

Check out this interesting greenscreen test. Which image will do best? 4k 8-bit 4:2:0, 1080 10-bit 4:2:2 or 2,5k 12-bit Raw?

 

[jgharding]

GH4 I reckon

 

you have more pixels to work with, so do your action shooting (or the whole thing) in 4K, do the post in 4K and deliver 1080p covered up with a bit of grain.

[Michael1]

Check out this interesting greenscreen test. Which image will do best? 4k 8-bit 4:2:0, 1080 10-bit 4:2:2 or 2,5k 12-bit Raw?

 

Very informative video.  The results were surprising.

 

Michael

[Jiban Huidrom]

I think it should be the best option !!!!!!!!!! BMPCC is now @ 495$ only

[Jiban Huidrom]

proress from bmpcc is amazing!!!!! No problem when keying.....

[varicam]

I think it should be the best option !!!!!!!!!! BMPCC is now @ 495$ only

 

A new model coming out?  This is an insane price.

[Rob Bannister]

This was always my issue with cameras is getting raw or 4:4:4 out of it. I went with the kinemini which is actually great but that pocket camera is so cheap now. I didnt like the sensor in it or the build but at that price who cares.

 

Just remember to light your green +1 stop to pull better keys especially when not shooting raw. Even prores and DnxHD get blocky if you go under. In film days is was always -1 stop but now we have to light brighter....which kinda sucks because you can get more spill but much better then trying to Roto fine details and Hair. Got bitten by that once on a show with the Alexa and never again.

[Rob Bannister]

Therefore, I stand behind my statement (and the statement of others on other forums- I didn't make up this theory), that a downsampled 4:2:0 or 4:2:2 image can be virtually indistinguishable from a native untouched 4:4:4 of the same size when it comes to grading and pulling a key. 

 

That's something else entirely. You're talking about frequency patterns, I think. 

 

Yeah the grids come from resampling the image...like anytime you do a transform you take a filter hit unless you are staying at whole numbers and not scaling or rotating this will be an impulse filter with no resampling. The problem with impulse is that you will see tearing if you do resize or rotate. So picking the best filter is necessary when doing these transforms, like in cubix you will start to see a grid pattern come out if you look close. 

 

Some info on filters taken from here http://www.fxguide.com/featured/Keeping_Your_Renders_Clean/

 

Nearest Neighbor/Impulse/Dirac: This algorithm samples only 1 pixel…the closest one to the original point. In the end, if you’re scaling upwards this simply ends up making the pixels bigger. If you’re doing an effect where you want the image to appear pixellated, this is the one for you. If you’re scaling down, the image will be very sharp yet pixels will boil and jump between frames in many situations.

Box: A 1×1 square is used for interpolation. As the name suggests, this gives a boxy look and in most cases is not useable for scaling up.

Bilinear: A 2×2 square is considered when interpolating imagery, ending up with a weighted average of the four pixels. This results in imagery which is far smoother than nearest neighbor.

Bicubic (“cubic†in Nuke): An even larger area, 4×4, is considered for bicubic interpolation. So you end sampling a total of 16 pixels. Pixels which are closer to the original point have more of an impact on the end result (aka “weighted†higher). Imagery is shaper than Bilinear, yet smoother than nearest neighbor.

Spline, Sinc, and others: Even larger areas are considered for these interpolations. They retain even more image detail than Bilinear or Bicubic when uprezzing or rotating, so if you’re doing multiple rotations which are not concatenated, they are generally a better way to go. However, these types of calculations are much more processor intensive and take a longer time to render. If you don’t need this extra detail — as mentioned above, this can depend upon your imagery — you might be able to get by with other interpolation methods.

Sinc: Keeps small details when scaling down with good aliasing. Ringing problems make it a questionable choice for scaling up. Default size is 4 x 4. It can also deliver negative values, which can be interesting when working in float/channel bit depth. It is one of the best methods for scaling down.

Mitchell / Catmulll-Rom : A good balance between sharpness and ringing, and so a good choice for scaling up. Sometimes Mitchell can produce better result with details, but it depends upon the content. Default size is 4 x 4.

Lanczos: Similar to the Sinc filter, but with less sharpness and ringing. Most likely better quality than Mitchell /Catmull-Rom.

Gaussian: Resizing is generally soft, but is good with ringing and aliasing.

Jinc (Autodesk Proprietary): Found in toxik, this resizing algorithm compares to a Lanczos filter in quality — with very good results maintaing details without softening.

[mtheory]

This was always my issue with cameras is getting raw or 4:4:4 out of it. I went with the kinemini which is actually great but that pocket camera is so cheap now. I didnt like the sensor in it or the build but at that price who cares.

 

Just remember to light your green +1 stop to pull better keys especially when not shooting raw. Even prores and DnxHD get blocky if you go under. In film days is was always -1 stop but now we have to light brighter....which kinda sucks because you can get more spill but much better then trying to Roto fine details and Hair. Got bitten by that once on a show with the Alexa and never again.

 

FxGuide did tests with the EPIC and also recommended overexposing for better keys.

 

Were you shooting at -1 stop or normal when you got into this issue with Alexa? 

 

Also, how is the rolling shutter at KineMINI 4K ? Do you ever get issues tracking handheld shots? 

[Rob Bannister]

Yes the issues came out when we told them on set to shoot a stop under. It pushed the codec too far and we were unaware that they were shooting directly to DnxHD and not to 444 DPX or raw. I think everyone was doing the change from something like the F35 to the Alexa. This was a few years ago it seemed everyone had this footage from episodic that you could not pull a clean key from.

 

It is a lot better if you expose your GS one stop up but there still is macro blocking in these codecs...you can just make dealing with it a bit better...true 444 or raw is still the way to go.