Shooting RAW - Most economical camera?

Can you elaborate on this? Do you mean, taking 4K footage from a camera and downrezing it to 2K or 1080p is a better idea than taking raw footage from the 5D Mark III or from the BMPCC? If this is true, would 4K footage from the Sony AX33 or Sony AX100 dowrezed to 2K or 1080p give footage good enough for chroma key work and color correction/grading?

Thanks.
High resolution has own advantages over a mere Raw recording. If one looks at the identical footage shot with the different resolution cameras - let's say, BMD Pocket Camera 2K Raw vs. GH4 4K 10-bit ProRes - many a person will prefer the 4K ProRes image. I know I would under most circumstances.

BTW, Digital Bolex is a prime example of that because I find its images far too soft.

And, if you use the green (chroma) screen, then 4K is almost a must.
 
Can you elaborate on this? Do you mean, taking 4K footage from a camera and downrezing it to 2K or 1080p is a better idea than taking raw footage from the 5D Mark III or from the BMPCC? If this is true, would 4K footage from the Sony AX33 or Sony AX100 dowrezed to 2K or 1080p give footage good enough for chroma key work and color correction/grading?

Thanks.

I haven't done extensive testing along these lines, but - two 4K keying schools of thought:

Properly downscale 4-2-0 4k to get 4-4-4 1080p and key away… (and still some arguments about what "proper" is);

Pull your key at 4K, and then either composite at 4K (dicey of you can't get plates and elements that big) or downscale your keyed footage and comp at 1080.

Anecdotally, option 2 is better - when it comes down to individual hairs and so on, you get more detail even though the some color channels are compromised.

But my guess would be that this may really vary shot-by-shot - just due to the almost ghost-in-the-machine nature that a lot of this work presents until you really dig in and test and figure out what's going on. With the availability of 4K still being fairly new, unless you shoot and comp affects all week long, we may still be building the best-practices procedures - anyone remember trying to key DV footage? After a couple years there were pretty standard workflows - convert the color space, etc.

I did a project a while back where every shot was keyed - something like 170 shots?? And it was shot with an HMC-150, which isn't a full-raster camera (the 1080 footage is upscaled from something a bit higher than 720). So test test test, and I ended up pulling keys at 1080 and comping and delivering at 720. Came out surprisingly well (most all the hair held up) but it took some playing around and the quality of the keys evolved over the month of post the project took.

The other thing that will affect the lone shooter/comper and the small studios with the guy who gets good at AE and gets all the comp work: being able to get 10 bit 422 4K prores or even raw footage, affordably. We're likely a year away from price drops in external 4K recorders (and they're already pretty affordable for anyone doing a fair amount of keying work) and it seems to me that the Ursa Mini could be a "last camera you need for years" for maybe two to three times the outlay for a Sony or Samsung and external recorder.

Last month Creative Cow sent an email update with a "keying tutorial" - and it was some kid in front of a popup screen and his hair looked like it was covered in meth-jacked crawling ants. So hopefully one thing we'll see with 4K is a lot less shitty keys (and "how to pul a **** key" tutorials) on youtube...
 
Looks to me OP is confusing raw with uncompressed video. 4:4:4 is a term that applies to debayered and YUV converted video, not to raw video.
Also, for chroma key work higher resolution will result in cleaner keys, so 4K will be better than a lower resolution camera, even if the target is HD.

There's always a nerd....lol Does it matter? :tongue:
 
... We're likely a year away from price drops in external 4K recorders (and they're already pretty affordable for anyone doing a fair amount of keying work) ...
Semi-off topic but --- two years ago, AJA Ki Pro Quad was retailing for $4,000, which was the lowest priced 4K external recorder (and not much of a monitor, with the whole thing weighing a hefty 4.5 pounds). These days, one can get a 4K external with a 5" monitor for as little as $1,200 and weighing under one pound.
 
+1 to what Michael Carter said.

A Samsung NX1 is probably the best camera on its price segment if your main task is to shoot greenscreen: it records very sharp 4K with very low noise.

And yes, knowing whether the OP wants RAW or 444 does matter, if you're trying to get him what he wants.
 
Is all raw not debayerable in any format to include 444

Ive always thought that (uncompressed) 444 was just like raw but less flexible colour wise and far more inefficient with data.

S
 
In terms of how useful it is, you're probably right. But for example if the in-camera debayer is not good, they are different.

In any case, they are different things (RAW means "not debayered", 444 means "debayered but with no color subsampling"), which makes the initial post a bit confusing.
 
When I started this thread, I did have some confusion about RAW and 4:4:4 footage. This has turned into a very informative discussion, nevertheless!

@Michael Carter

Thank you for your elaborate explanation. I have been recommended a Sony A7S on rent for a project for 2k/4K footage. I have been advised that this will be suitable for chroma key work as well as color correction and grading. I will read up on it in older threads! Thanks.
 
High resolution has own advantages over a mere Raw recording. If one looks at the identical footage shot with the different resolution cameras - let's say, BMD Pocket Camera 2K Raw vs. GH4 4K 10-bit ProRes - many a person will prefer the 4K ProRes image. I know I would under most circumstances.

I can get a better greenscreen key with 5d raw than 4k from an a7s or somesuch with huge compression artifacts.
 
In any case, they are different things (RAW means "not debayered", 444 means "debayered but with no color subsampling"), which makes the initial post a bit confusing.

So if you shoot raw you can always get 444 right? So a shooter who needs 444 must shoot 444 or raw.

--

To the OP.

I think any 420 onboard codec will not work for you. Maybe there are offbord recording options for the a7 cameras - I dont follow offboard recorders because I dont want my client data going down a flimsy wire.. especially HDMI.

BMC cameras have some issues in the field with no ND or sound and somewhat limited dynamic range but in a studio they would seem to be the correct option, the 2.5, the 4k or the ursa will produce footage that has far more colour fidelity than any mirrorless/dlsr which is not using its full sensor read to make video.

Whatever specs you read Im sure that BMC files have a lot more heft than mirrorless files.

S
 
So if you shoot raw you can always get 444 right? So a shooter who needs 444 must shoot 444 or raw.
The short answer is "yes", but it's a nuanced topic.

If you have a 1920x1080 Bayer pattern, you can get a 1920x1080 RGB signal, or a YUV-444 signal, but part of that color information has to be interpolated because the original Bayer pattern only reads one color at each photosite. Some people say this means that from 1920x1080 Bayer you can only get 422 or even 420 color, but debayering algorithms today include edge detection and other advanced features, meaning you can almost get back the full 444 color; not quite, but close. And in any case, the type of color-information-loss from the debayering is different from the type of color-information-loss from the color subsampling, so storing your debayered image in 444 will preserve everything you can get out of the sensor, whereas going to 422 or 420 will kill some color information, on top of what you already lost because it's a Bayer pattern and not a Foveon-like sensor, or a triple-CCD design, or an oversampled Bayer (C300-like), etc.

In the end, details such as the codec or the noise floor are often more important than all this. The only way to know if the NX1 (shooting 4K 420 H.265) or the 5D3 (shooting 1080p RAW) will give you a better key is to test them side by side (even a zone plate test chart is not good enough, because it doesn't show the effect of compression when there's motion).
 
Theoretically speaking, any resolution will give you 444 at a much lower resolution. 4K will deliver 444 at 1080p (although, with the 4:2:2 being nearly identical to 444 with most modern algorithms, 2.5K 4:2:2 should be a better approximation ... not that many clients ever ask for 2.5K delivery). 1080p in 4:2:0 will deliver a 444 at about 650 lines (or whatever adds up to ~ 700 KPX). This is why top tier cameras - and A7RII and DVX200 - will oversample with their pixel count achieve both the resolution and the color accuracy, as the two are a trade-off when deploying a Bayer type filter.
 
Back
Top