ProRes 4444 - DSLR usage

8salacious9 · Dec 7, 2012

So I'm trying to write a short guide for dslr film makers who are just getting into it.

The problem I have at the moment is with ProRes 4444, there's something which I cannot explain or don't understand if I'm honest.

ProRes 4444 includes an alpha channel, now I want to present scenarios which I want you to say its correct for, as I'm still trying to clarify its use in a workflow.

1. You have shot some green screen and you need to preserve the information for more accurate keying.

2. You have shot normal footage and would like to apply some vfx via after effects or another graphic programme.

3. Both of the above

From what I have researched it would seem the alpha channel is needed only if you need to key out some colour? So why would this not work with 422 codec? Or if it does is it not effective enough?

Thanks

jax_rox · Dec 7, 2012

This is what it's all about:
http://en.wikipedia.org/wiki/Chroma_subsampling

ProRes 422 is 4:2:2 chroma subsampling, which is generally the accepted standard for broadcast. ProRes 4444 samples at 4:4:4 which is the highest quality you can get, and is essentially RGB, which is why you'd use it. It also has a 12-bit depth, rather than 422's 10-bit. That's why it's used a lot on high end cinema production, though television productions and TVCs use it a lot now too. It also gives you the extra alpha channel support.

Re the alpha channel, you want that if you're transferring footage between computers. Yes, you could key 422 footage, but if you're transferring that footage around, the alpha channel will get lost. With 4444, you can keep the alpha channel with it. Generally, however, that's not the main reason for it's use, it just can be handy to have. Usually 4444 is used when you want as much colour information retention as possible, ie cinema/film applications etc.

8salacious9 · Dec 7, 2012

jax_rox said:
This is what it's all about:
http://en.wikipedia.org/wiki/Chroma_subsampling

ProRes 422 is 4:2:2 chroma subsampling, which is generally the accepted standard for broadcast. ProRes 4444 samples at 4:4:4 which is the highest quality you can get, and is essentially RGB, which is why you'd use it. It also has a 12-bit depth, rather than 422's 10-bit. That's why it's used a lot on high end cinema production, though television productions and TVCs use it a lot now too. It also gives you the extra alpha channel support.

Re the alpha channel, you want that if you're transferring footage between computers. Yes, you could key 422 footage, but if you're transferring that footage around, the alpha channel will get lost. With 4444, you can keep the alpha channel with it. Generally, however, that's not the main reason for it's use, it just can be handy to have. Usually 4444 is used when you want as much colour information retention as possible, ie cinema/film applications etc.

and with that, the info i shall put is, for beginners, this isnt for you

cheers for that

ItDonnedOnMe · Dec 7, 2012

The other thing to consider is that with a DSLR you're generally shooting 4:2:0 in camera. Transcoding to 4:2:2 will preserve all the color info the camera stores - but you don't gain anything by going to 4:4:4:4 except larger file sizes, so it's not something you'd benefit from until later in the post process.

So yes, not something for beginners.

knightly · Dec 7, 2012

The alpha can contain depth pass as well for 3d compositing on the way out of the renderer.

jax_rox · Dec 8, 2012

ItDonnedOnMe said:
The other thing to consider is that with a DSLR you're generally shooting 4:2:0 in camera. Transcoding to 4:2:2 will preserve all the color info the camera stores - but you don't gain anything by going to 4:4:4:4 except larger file sizes

Well, you don't gain anything by transcoding to 4:2:2. 4444 preserves as much as 422 does, just that 4444 will give you an unnecessarily larger file size for no extra gain.

Similarly, you could be safe with using DNxHD 36, which is Avid's 8-bit, 36 Mbps intermediate codec, as a 5DmkII compresses at around 38 Mbps. Transcoding to say, DNxHD 220, giving you 10-bit, 220 Mbps, would be pointless as there'd be no gain in picture quality.
Shooting on a RED, however, and treanscoding to DNxHD 36 vs 220, or ProRess 422 vs 4444, you would notice a difference.

8salacious9 · Dec 8, 2012

My initial investigation was the conviction to use 4444 because of the alpha channel.

ItDonnedOnMe · Dec 9, 2012

jax_rox said:
Well, you don't gain anything by transcoding to 4:2:2. 4444 preserves as much as 422 does, just that 4444 will give you an unnecessarily larger file size for no extra gain.

That's pretty much exactly what I said?

8salacious9 said:
My initial investigation was the conviction to use 4444 because of the alpha channel.

It depends a lot on what part of the post workflow you're talking about - I was assuming you were talking about which format you might convert the camera originals to for editing. There's no alpha channel coming from the camera, so there's no need to convert your camera footage to a format which supports one. Now if you take your footage into AE, key or mask part of it, then want to take it into another program while preserving that transparency - at that point you might render those clips out to something like ProRes 4444 to get the alpha channel.