How to get file with Alpha into ProRes422 timeline?

Working on a program in FCP 6.0.6. ProRes 422 (non HQ). Creating CG's in Motion.

I'm working on some lower-thirds for the program in Motion, and want to pull them into FCP. Running into an issue to which I'm sure there's a solution, but haven't found it yet myself, so thought I'd ask here. It seems that ProRes 422 can't deal with alpha channels? I've found conflicting reports about this, so not sure whether true or not. Seems odd that a relatively new codec wouldn't be able to handle alpha's...

I've tried exporting from Motion in several ways (with Alpha channel, Alpha + Color, Color only, using ProRes 422 and DVCPROHD) and messing with blending modes, but can't seem to get it to show up as anything other than my lower third with the rest of the screen black (i.e. alpha not being properly interpreted). The motion file will drop into FCP and seems to render just fine, but I've had enough trouble doing that without Render glitches, that I've been exporting everything from Motion to a quicktime before FCP import. So the direct import is an OK temp. solution, but I'd like to figure out something that allows me to export a quicktime before FCP import.

Any ideas? Hopefully I'm just missing the totally obvious solution. Thanks, folks.

You have to export a separate animated "travelling" matte if you use ProRes 422 as very few video CODECs have built-in Alpha Channels.

However ProRes 4444 Does.

However ProRes 4444 was released with Final Cut Studio 3 and FCP7.


The options for you are:

A. To export a PNG or TIFF sequence with transparency or Animation CODEC with Alpha.

B. Create an Alpha channel matte (static or travelling) for your video "overlay" to cut it out.

or the best option

C. Import the Motion Project File DIRECTLY onto the timeline with transparent background and render in place.



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

thanks, Ben, looks like I'll be using the motion direct to FCP method on this one, as long as I can get it to render without issue.

On a related note, in the DSLR footage world, why is it that the h264 to ProRes 422 conversion is so heavily pushed? I know that FCP can't natively edit h264, but why wouldn't people recommend ProRes 4444 or DVCPROHD as an acceptable format, especially if they don't exhibit the same limitations as ProRes 422?

Prores 4444 is an RGB codec, so it may clip whites and blacks.

Prores/prores hq is perfectly fine for canon DSLRs. Prores is designed as an intermediate codec, to allow you to work with temporally compressed formats, while still able to preserve good video quality across multiple generations. Also, you don't need the additional 0:2:2 chroma samples, as your acquisition codec is 4:2:0.

You have 3 kinds of codecs- production or acquisition codecs, post production or intermediate codecs and delivery codecs. They all serve different functions. Eg. On location, you want smaller file sizes, and something which can allow for efficient compression with hardware acceleration (like h.264 or hdv or RED), in post production, you want a codec that allows for more real time effects, semi Quasi transparency across generations, and you have higher data access rates and more storage, while for delivery, you want something that specific consumer devices can decode. So they are all designed for different purposes. Some codecs like dvcpro hd was designed mid way between an acquisition codec and an intermediate codec, so it fits a full HD file into 100 Mb/s without long GOP encoding.

What Ben said. Either export an animation codec with alpha channel embedded or export the alpha channel as an alpha matte (black and white image that represents your alpha values).



www.strypesinpost.com

strypes Wrote:
-------------------------------------------------------
> Prores 4444 is an RGB codec, so it may clip whites
> and blacks.

It supports Y'CbCr 4:4:4, as well, so it shouldn't "hurt" anything.

Quote
Apple ProRes White Paper (July 2009)
Apple ProRes 4444: The new Apple ProRes 4444 codec preserves motion image
sequences originating in either 4:4:4 RGB or 4:4:4 Y?CBCR color spaces. At its remarkably
low data rate as compared to uncompressed 4:4:4 HD, Apple ProRes 4444 supports
12-bit pixel depth with an optional, mathematically lossless alpha channel for true
4:4:4:4 support. Apple ProRes 4444 preserves visual quality at the same high level as
does Apple ProRes 422 (HQ), but for 4:4:4 image sources, which can carry the highest
possible color detail.

Or, am I missing something?


-Dave

Sorry I decided to do a brain fart...

Quote

looks like I'll be using the motion direct to FCP method on this one, as long as I can get it to render without issue.

Yep rendering Motion Projects can be a bit flakey, but if it works it should give you the functionality and ease of alteration via simply changing the Motion Project and re-saving it.

Quote

why is it that the h264 to ProRes 422 conversion is so heavily pushed? I know that FCP can't natively edit h264, but why wouldn't people recommend ProRes 4444 or DVCPROHD as an acceptable format

H264 (MPEG4) in such a low datarate was never designed to be an acquisition format; as the amount of processing it needs to compress and decompress is insane - also the quality is dreadful at low bitrates.

The main reason we say change to ProRes is partially the overheads for processing are much less than for H264.

Also any grading or any additional work added to the timeline wont be heavily compressed into the H264 stream (although you might want that later for web delivery).

ProRes 4444 is not needed by most people as it's over and above what you deliver for broadcast; the top specifications I currently work to for BBC et al - are only 10bit 422 for both HD and SD.

But ProRes 4444 IS useful for its Alpha Channel in any workflow that needs one.


4:2:2 video is an odd format as it is not made up of Red, Green or Blue channels (which we see on our Computer Monitors and how we create Graphics) but merely a different way of encoding RGB information that we do eventually see on our TV screens as a mixture of Red, Green and Blue light.

But 4:2:2 does refer to 3 channels (A:B:C) of picture information:

A. one Luminance channel sampled at full resolution

B. chroma channel that is sampled at half the resolution of the Luminance.

C. another chroma channel that is sampled at half the resolution of the Luminance.

So a 640x480 4:2:2 image would have 640x480 pixels for luminance and then one chroma channel at 320x480 and the third channel at 320x480. These then are mixed together to form your video image.

This is because the human visual system is much more sensitive to variations in brightness than in colour and so the early video engineers thought "hell we can throw away loads of the colour information in a picture and no-one will notice!". Which no-one did and video signals were henceforth compressed.

So now in the digital age we have sampling like 4:2:2, 4:2:1, 4:1:1, and 4:2:0 because we know for a lot of applications we can get rid of, or heavily compress the colour.

By the way, your DSLR only captures 8bit 4:2:0 which throws away more of the colour detail than a 4:2:2 camera.



When I say 8bit or 10bit this simply defines the amount of samples for the levels or the value of each pixel in each channel - it has nothing to do with the size of the pixels in the samples.


To understand a bit about bits (pun was almost avoided and then left in)

A binary bit has one of 2 values

either OFF or ON

0 or 1

So if we say '8bit' we mean 2 to the power of 8


On your Calculator App...

If you press 2 then the multiply button you can step through the possible bit values by pressing equals

Starting at 2 or 1bit

press equals and you get the value 4 or 2bit

Press equals again and you get 8 or 3bit

and again you get 16 or 4bit, etc...

Each time doubling the previous number.


So to 8bit or 2 to the power of 8... or 256


8bit gives you 256 discreet levels of what ever you apply it to...

Such as 256 different text characters in the standard 8bit ASCII character set

or 256 different levels or steps in 1 sample of digital audio.

or 256 different levels of colour in ONE of the sliders on an RGB scale



If we take the RGB slider example:

You get 3 sliders

1 for Red, 1 for Green and 1 for Blue

We can work out how many different combinations we can have:

256 x 256 x 256 = 16,777,216 colours

Or another way of looking at it:

8bit+8bit+8bit gives a total of 24bits

2 to the power of 24 is 16,777,216

Which is why 8bit per channel RGB video is sometimes referred to as 24bit.


Lets scale that up:

10bit gives you 1024 discreet levels

So our 10bit per channel RGB slider now has a total of 30bits (R10bit+G10bit+B10bit) or 1,073,741,824 possibilities

Which is why 10bit per channel RGB video is sometimes referred to as 30bit

16bit gives you 65536 discreet levels - our 16bit per channel RGB slider now has a whopping 281,474,976,710,656 levels... aka 48bit video.



So 10bit video allows for 1024 levels of gradation in each of the channels meaning smoother colour/luminance and little no horrid banding of gradients!

--------------

Back to the question:

DVCproHD is recommended but its only 8bit 422 - and we all (broadcast fcp'ers that is) pretty much decamped to ProRes 422 because it also allows for 10bit video which again allows for better gradient scale in the luminance and colour but it is also a lot less demanding of storage space than Uncompressed 422 10bit Video.

--------------


Apologies for the mental evacuation... (or any over simplification for those [Graeme] reading the explanation)

If you read this Video compression and sub-sampling might be made a bit (yes very punny) clearer: [en.wikipedia.org]



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

>DVCproHD is recommended but its only 8bit 422

Also, DvcproHD subsamples. In PAL frame rates, the frame size of Dvcpro HD is 1440x1080, and in NTSC frame rates, it is 1280x1080. That is a LOT of resolution lost.

>in either 4:4:4 RGB or 4:4:4 Y?CBCR color spaces.

Shane once mentioned about clipping on ProRes 4444 converted material. I'm not sure how ProRes figures if it's Y'CbCr (A) or RGBA or when to use either. In theory, Y'CbCr exists as a 4:4:4 format, and HDCAM SR supports it, but there is no reason to decimate your chroma values.

The difference between an RGB and a YUV codec is where it clips. On a YUV to RGB conversion, video black point is usually mapped to absolute black for RGB, while video white is mapped to absolute white in RGB. If you have information in the sub blacks or in the superwhites, that would get clipped. I really should do more tests on this and whether ProRes can intelligently map data above and below black/white points.

I haven't used ProRes 4444 extensively myself, but there's no usually reason for me to go to ProRes 4444, as I usually deal with video. When I work with RED, yes, and no. Sure, RED is a 4:4:4 RGB source, but I usually end up going to tape for broadcast, which is 4:2:2, so HQ is sufficient. When going to the online machines, I tend to debayer to 16 bit tiffs.



www.strypesinpost.com

Hey guys try to use Y?CbCr not YUV

YUV is Analogue



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

Sorry, it's just easier to type, and the assumption isn't too hard to make.



www.strypesinpost.com

lol



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

strypes Wrote:
-------------------------------------------------------
> Sorry, it's just easier to type, and the
> assumption isn't too hard to make.

It is confusing, especially for those who don't know the differences involved. I agree with Ben that we should use proper terminology whenever possible.

See this: YUV and luminance considered harmful:
A plea for precise terminology in video.



-Dave

Harmful or not, I am not typing it out on the phone!



www.strypesinpost.com

strypes Wrote:
-------------------------------------------------------
> Harmful or not, I am not typing it out on the
> phone!

3 extra letters... how about trying this: TextExpander touch?




-Dave

2vuy, v210, ARGB...



www.strypesinpost.com

I can't ever remember Y apostrophe c..crB .. something.. but, I do love little kittens...

strypes Wrote:
-------------------------------------------------------
> 2vuy, v210, ARGB...

No need to get medieval on us...

Jude Cotter Wrote:
-------------------------------------------------------
> I can't ever remember Y apostrophe c..crB ..
> something.. but, I do love little kittens...
>
>



I hate to admit it, but I, too, love little kittens and puppies...


Thanks for that, Jude.


-Dave

Damned. I can't watch that. It's.... Flash!



www.strypesinpost.com

OMG! That was hilarious!

Back to when you use ProRes and when you use ProRes 4444... When your source footage is a long GOP video source and you are delivering for a video output, you use ProRes. The difference between ProRes and ProRes HQ is in the aggression of compression. You will lose a little bit of quality during the transcoding process, but it lets you work more efficiently in post production, also, you can master to ProRes, whereas, you never want to use H.264 as a mastering codec because is not good enough as an intermediate codec.

ProRes 4444 is designed for use with high resolution formats such as RED which shoots to 4:4:4, and also when you need an alpha channel. You can also use ProRes 4444 on compressed video sources, especially if it does not clip black/white levels, but you gain nothing in doing so if your source is 4:2:2, 4:2:0, or 4:1:1, since none of those formats will use that extra 0:2:2 chroma subsamples.



www.strypesinpost.com

for motion graphics with quick and dirty built in alpha I use the Animation codec. But the results will vary depending on the complexity of the graphic. Sometimes there's no way around using motion mattes that match the graphic. You would need to make a black and white version of the exact file with white representing everything you want to see and black representing whats invisible. You put the B&W version on track two and the full color version on track three, then set the track three clip composite mode to Travel Matte Luma. This way you can export both your color graphic and B&W matte as prores.