444 Codec & Video Processing

[Gary Adcock]

[Rivai Chen] “FCP can only set to high-precision YUV or RGB 8 bit.”

There is nothing that I can find published by Apple that say that FCP is limited to 8bit color, many of the filters and effects -yes- but not the processing. As of FCP 7’s release RGB material can access 32-bit float space for rendering. (do a search for 10bit RGB in FCP help)

While a single link HDSDI signal is limited to RGB8 bit, a Dual Link signal easily supports up to 12 bit log as RGB so I do not understand where that thinking is from. Maybe its because the RT engine in FCP defaults to 8bit for rendering as do most of the filters and effect.

If you are working in 10bit RGB RT is turned off by default, there is a reason when there is no realtime processing any bit depth in the original is handled without issue.

“this is FCP that we’re discussing here. i find there’s no way to have such setting RGB 10bit even if you have the AJA hardware.”
See above.
Here in lies my last point in the previous post- for this type of highend workflow FCP is traditionally only part of the picture, and that is why there are specific easy setups for aja’s 10bit log codec.

“The original content is comes from RED 12bit linear and it’s RAW. “
Its only 12bit if you handle it correctly- that would assume that you are converting to DPX within the RED software or using one of the highend tools that support R3D output at that quality.

“I would like very much to maintain it until the final export to DPX via gluetools”
Why not convert your content to DPX to start with?

” If any link in the chain can’t handle FP then the whole chain drops back to 8 bit. In our experience few workflows through FCP are greater than 8 bits, “

I know David and he is one smart guy and in Cineform they have the single best cross-platform codec available. He is correct when he says that few have workflows that are above 8bit.

In this quote he never says, as you imply, that it cannot be done.

Many of the FCP filters and effects render in 8bit and the RT engine for the most part working in 8bit, but within ProRes it does maintain 10bit processing of YUV. Unfortunately Apple does not allow other codecs (like Cineform or Sheer) that privilege. Turning off the RT engine does allow you the ability to handle the content at the native bit depth with less interference.

Working in 10bit RGB within FCP is just not as common as people think it is and it takes diligence throughout the process, thats why the highend has always been about the proprietary tools and processing and workflow.

gary adcock
Studio37
HD & Film Consultation
Post and Production Workflows for the Digitally Inclined
Chicago, IL

https://blogs.creativecow.net/24640

[Rafael Amador]

Hi Rivai,
There is nothing published by Apple, neither inside FC, that leads to think that 10b RGB rendering is possible.
In his post, the guy of CineForm, basically says that they are unable. They have to convert and render in YUV.
Last year, after similar discussion, I asked Graeme Nattress about. Also Owen Smithyman of BitJazz (Sheer codec).
Nobody even considered the possibility of FC rendering in 10b.
Owen proved that was possible to cut and export without rendering, keeping the 10b RGB.
Personally, I don’t understand any piece of software hiding what should be his best highlight.
Cheers,
Rafael
PS: 444 is not a color space but a “mapping/sampling pattern” (422, 420, 411).
Tolking about 444RGB makes no much sense as long as there are no 422/420/411 RGB formats.
Down sampling in RGB is not possible.

http://www.nagavideo.com

[Rivai Chen]

Hi Rafael,

Yeah agree, the most important note from David Taylor is the calculation between YUV and RGB in FCP is done thru floating point calculation. That’s the key. As long your incoming and destination format can handle this, it should be alright. This is ensuring enough. I still have no idea why this is not documented. Sometimes, we just need to ask developers about the factual data. There’s other misleading or half complete information spreading around that QT for windows is always 8 bit and designed that way which of course not the case. I heard such false assumption from other people who supposed to be pro. The point is sometimes we have to do a lot of check and asking around to seek such information, it will be so much easier it’s documented properly.

Anyway, at least i got the answer 🙂

Rivai

[Gary Adcock]

[Rivai Chen] “There’s other misleading or half complete information spreading around that QT for windows is always 8 bit and designed that way which of course not the case”

The QT player application is 8bit on both mac and windows. The files may not be, but the base app is.

gary adcock
Studio37
HD & Film Consultation
Post and Production Workflows for the Digitally Inclined
Chicago, IL

https://blogs.creativecow.net/24640

[Rivai Chen]

Which is exactly half complete and people forget there’s a deep pixel precision.

[Rafael Amador]

[Rivai Chen] “Yeah agree, the most important note from David Taylor is the calculation between YUV and RGB in FCP is done thru floating point calculation. That’s the key. As long your incoming and destination format can handle this, it should be alright. “
No, this is not alright, but it seems that is the only chance.
You don’t go RGB>YUV>RGB, when you can stay in RGB all the way through.

[Rivai Chen] “I still have no idea why this is not documented. Sometimes, we just need to ask developers about the factual data. There’s other misleading or half complete information spreading around that QT for windows is always 8 bit and designed that way which of course not the case. I heard such false assumption from other people who supposed to be pro.”
Analog was more difficult to manage, but everything was standard. Everything was in the books.
In NL digital, everything (Applications, codecs) are proprietary. They do whatever they want inside their software and they don’t have to document it.
Professional Third Part developers complains that they are absolutely lost about many internal processes in FC.
Some says that Apple do not disclose documents because there are too may internal tricks and substandard processes used to overcome the FC and QT shortcomings.
rafael

http://www.nagavideo.com

[Uli Kunkel]

Gary,

The documentation in FCP 7’s manual on 10 bit RGB processing is vague and non-specific from what I can tell. Everything you’ve said looks to be true, but I’m still looking for verification on a couple of things.

– If an FCP Timeline set up for ProRes 444 is set to “Always Render RGB,” will I still be outputting 10 bit RGB color through my Kona card onto SR?

– Will simple dissolves render to 10 bit RGB in the timeline, or will they only be 8 bit?

Somewhat unrelated question…

– I am rendering Red footage from Color to ProRes 4444 for a Davinci 2K color correction off of SR tape. If my FCP timeline is relinked to the Red _H Proxies and sent to Color, will Color be accessing the same color space (12 bit RGB) as it would if I had used the “Native” workflow from the Red FCP whitepaper?

Thanks for your advice.

Uli

[Eliot Piltz]

Rather than start a new topic, I have a related question. Can someone advise me on the render settings for a final output of a show I’m finishing? It was edited at ProRes422HQ, graded in Apple Color and rendered out in 16-bit ProRes4444. Now back in FCP, there are many titles, subtitles, transitions, speed changes etc. that have to be rendered. Should I set the video processing to high-precision YUV or RGB? This is going to HDCAM SR for festivals, but of course should be broadcast safe, as we only want to make one master. Also, is there any need to change the ProRes gamma settings to None, or will “Automatic” not mess anything up?

Thanks very much!
Eliot

Mac Pro 2.66 GHz 8-core (4,1); 12 GB Memory
ATI Radeon HD 4870 (512 MB); ACD 30″ and 24″ dual display
Sony BVM-L230, Blackmagic Intensity Pro, Drobo 8-bay 16 TB
OS X 10.6.2, FCP 7.0.2

[Johnathan Throbins]

So what are you guys using to online?
It smoke the only answer if you want to retain a 10+bit RGB workflow from Resolve? Do you have to online in resolve?

In particular when I deliver a feature sometimes the original editor wants to add the titles or someone will crop from 16:9 to 2.39 for a DCP and render that, I’m worried they’re bumping it down to 8bit.

Also Resolve isn’t good with the editing, even in resolve 11 I don’t love it. I’d rather go back to FCP 7 but that’s not a great option. Is there an affordable online editing tool? Is smoke the cheapest good option?

Thanks.