Apple Color REDCODE plugin 4.0 beta

[Luis Otero]

Graeme's opinion,

Graeme's opinion,

Since I have read and heard that in fact you are baking in data when transcoding, I still do not understand why is better to transcode? The FCS Whitepaper outlines the folowing advantage:

Advantage —Using an EDL and Cinema Tools database to move your sequence into Color and relink it directly to the native REDCODE media eliminates the added step of reingesting using the Log and Transfer window. Relinking using a Cinema Tools database is fast: you can relink to media on multiple hard drives in one step. Working with REDCODE media provides all the same advantages as working with native RED QuickTime files.

This saves me time, as well as assure me that, as cited above, I am accessing the R3D, hence better results should be obtained. Is there a hole on this logic?

Please advise,

Luis


PS:

I am not arguing, I just want to have all the information on hand before changing all the workflow I have always used based on the literature available.

 

[Luis Otero]

It does indeed work "fine" but you are missing out on the greatness inherent in your footage (sharpness / detail + FLUT™™ and new color science)...
- Jordan

Jordan,

The new Color Plug-in contains the new color science. I can select the Gamma and the Space. Not FLUT yet, but... Still confused...

 

[Graeme Nattress]

Sharpness issue is an interesting one. Could very well be that the rapid 2k decode from 4k looks sharper as it has no downsample filter in place. However, 4k has more detail when done to full demosaic, it just needs a small bit of post sharpening as necessary to make it pop.

New Color Plugin does have the new CS in it, and most of the benefits of FLUT - just not the FLUT control for fine control over ISO.

Native through FCP / Color has worked well for me, but I only ever shoot home movies

Graeme

 

[Luis Otero]

Thanks for your input! Greatly appreciated.

 

[Jordan Livingston]

Jordan,

Thanks for your response. However, since the proxies allow me to have the RED tab enabled, why would it not give me the best of the best, due to such flexibility? If the footage is transcoded, my understanding is that the RED tab cannot be enabled. Am I missing something here? Can you clarify, please?

Unfortunately, when you use the RED tab in Color, although you have access to your 'RAW,' you are relying on the Apple debayer codec, which at this time is still inferior to REDCINE-X's. Especially in terms of sharpness, the resulting images appear to be about a half-res debayer at best, and even that's questionable.

When you transcode in REDCINE-X to a robust codec (like ProRes444) using REDColor / REDGamma, the truth is, your transcodes retain enough lattitude to obtain excellent results in Color, plus the benefit of the superior debayer, plus the benefit of FLUT control. After tons of personal testing and consulting with other pros on this, I have yet to see any benefit to grading RAW in Color, other than perhaps an up-front time savings, which again, is also a significant quality compromise.

- Jordan

- Jordan

 

[Jordan Livingston]

Jordan,

The new Color Plug-in contains the new color science. I can select the Gamma and the Space. Not FLUT™ yet, but... Still confused...

Still has the [lack of] sharpness issue, plus no FLUT. My opinion is that, if you shot RED, you deserve to see the maximum detail / quality in the images that you captured.

- Jordan

 

[Alexander Alexandrov]

slightly digressing:
so in the case when not grading RAW in Color, what workflow do you use? transcode to prores4444 from inception cut it and color it? no re-conform? curious if the edit with large prores4444s is fast on a decent macpro.

 

[Jordan Livingston]

slightly digressing:
so in the case when not grading RAW in Color, what workflow do you use? transcode to prores4444 from inception cut it and color it? no re-conform? curious if the edit with large prores4444s is fast on a decent macpro.

I have REDRocket so transcodes happen in realtime. I transcode to ProRes4444 w / REDColor + REDGamma and sometimes whitebalance and FLUT adjustments, as needed, in REDCine-X.

If project is finishing bigger than HD, I do "online" later in 2K or 4K, as needed, re-transcoding selects only. For all HD-quality deliverables, however, the ProRes4444s work just great on MacPro 8-Core Nehalem w/ Kona LHi.

- Jordan

 

[Ben Brainerd]

Unfortunately, when you use the RED tab in Color, although you have access to your 'RAW,' you are relying on the Apple debayer codec, which at this time is still inferior to REDCINE-X's. Especially in terms of sharpness, the resulting images appear to be about a half-res debayer at best, and even that's questionable.

IIRC, Color is using the *exact same* debayer as the RED tools, based on the SDK. There was a big discussion about it back in the day. So it's not the debayer engine that's causing images in Color to appear not as sharp.

If I were to hazard a guess at what's going on under the hood, I would say that the RED tools are applying a certain amount of sharpening to the image, based on what "looks good". Kind of like how Lightroom applies a certain amount of sharpening to stills by default, since Bayer RAW images need it, to a certain degree. Whereas Color appears to be adding, well, none.

Personally, I prefer the Color implementation in a way. If I'm going to do a sharpening pass, which I almost always am, I prefer to have total control over it. I've managed to produce images out of Color that are effectively indistinguishable from RC-X images, so it's not like it can't be done.

 

[Gunleik Groven]

What is up for testing is A/B on signal chain.

The debayer is the same, as it is REDs.
The problem with Color has been more of a signal chain thing issue of wether the grade is basically done through adressing the metadata with extensive controll, or if the signal is locked to RGB out of the RED room, AND whether there is a full debayer or a half happening when going to 2k/1080.

A/B testing time...

But itlooks better than before...

 

[Ian Andolina]

So Graeme, your hypothesis is that the full 4k debayer in Color then uses a downsample filter (for example to a 2K out) that results in a softer image, but with more possible detail to sharpen than the faster 2K debayer in Redcine-X? Doesn't the fact that the OLPF / sharpening happening at the debayer stage in Redcine-X, where as it happens to RGB data much later in the Color pixel pipeline offset this potential advantage?

Are the exact denoise/OLPF compensation values used by the Red Color plugin fixed or is there any way we can tweak them (hack the plugin framework for example)?

 

[Julio Quintana]

Can someone confirm if this works with FCS2, or is it limited to studio 3?

 

[Ben Brainerd]

The problem with Color has been more of a signal chain thing issue of wether the grade is basically done through adressing the metadata with extensive controll, or if the signal is locked to RGB out of the RED room...

I'm pretty sure Graeme said earlier that the image is more or less handed off to Color after the RED tab. So I'm guessing that means it's RGB past the basic settings.

That's one advantage RC-X has over Color: FLUT and the like run down the entire image chain. It's just a shame RC-X isn't more 'grading friendly'.

 

[Gunleik Groven]

yup.

looks better now.

As I said. Need 2 test.

00000000000

 

[Brandon Kraemer]

yup.

looks better now.

As I said. Need 2 test.

00000000000

So... the Color renders scaled to 2k or 1080 look as good as Redcine transcodes with the same ISO characteristics, Color space and Gamma space?

or.. the new Color plugin looks better than the old one?

talking about sharpness of course.

 

[Gunleik Groven]

So... the Color renders scaled to 2k or 1080 look as good as Redcine transcodes with the same ISO characteristics, Color space and Gamma space?

or.. the new Color plugin looks better than the old one?

talking about sharpness of course.

Can't answer before testing, but from the pure initial "feel" it "feels" better, but that doesn't really make much sense, does it...

As to sharpening, I try to avoid it till the very end of a signal chain for a number of reasons, and cannot really have an opinion about that... It would be nice to have more controll and options in the Color debayer and scaling routines - I agree totally with that statement.

As to sharpening - I guess we have different preferences.

Sharpening early on creates so many additional challenges....

But again, that's just me...

 

[Julio Quintana]

Rob,
Is this plugin designed for FCS 2 or only Studio 3?

 

[Jordan Livingston]

IIRC, Color is using the *exact same* debayer as the RED tools, based on the SDK. There was a big discussion about it back in the day. So it's not the debayer engine that's causing images in Color to appear not as sharp.

Thank you for clarifying! This is the first time I'm hearing this news, which is surprising, given the results from the two apps are so visually different. So perhaps the 'problem' is not the debayer itself...

That said, I can't see how the debayer engine could be "exactly the same," since Final Cut Pro and Color are governed by the REDCode system preference panel, which, as you know, tops out at "Half High" being the maximum quality option. Moreover, even the "Half High" results coming from Final Cut Pro and Color don't rival the "Half High" results out of REDCine-X in terms of sharpness and detail.

If I were to hazard a guess at what's going on under the hood, I would say that the RED tools are applying a certain amount of sharpening to the image, based on what "looks good". Kind of like how Lightroom applies a certain amount of sharpening to stills by default, since Bayer RAW images need it, to a certain degree. Whereas Color appears to be adding, well, none.

This is a sound guess. For all I know, it's correct. But it's important to point out that the REDCine-X results, as you just said, "looks good," and with the REDRocket card those results are also realtime. I don't want to wait longer for renders from Color that ultimately look 'worse.'

Personally, I prefer the Color implementation in a way. If I'm going to do a sharpening pass, which I almost always am, I prefer to have total control over it. I've managed to produce images out of Color that are effectively indistinguishable from RC-X images, so it's not like it can't be done.

While it's true there are many methods of post sharpening, Color itself does not offer any "sharpness" or "detail" beyond what's found in the FX Room, which I believe applies processing AFTER the debayer, not actually to the RAW directly (I could be wrong here, but that's what it seems like). The sharpening FX plugin itself is not bad, but I'm wary of sharpening happening after the debayer / demosaic. I'd rather it was happening to the RAW first.

At the end of the day, you might be right that you can get better results out of the plug-ins with Color. It's also true that the debayering in Color is likely to improve dramatically with future software updates. However, my goal here is to be able to recommend a rock-solid workflow that works reliably today.

My clients are all precariously juggling time, quality, and of course, money. I have personally found that the best quality in the least amount of time and money comes from realtime dabayering to the deliverable size / ratio (i.e. 1920x1080 HD in ProRes4444) first, then color grading in REDColor / REDGamma (or REDColor / REDLog) in Color. This workflow has never yielded any complaints about sharpness and has always finished smoothly, without any 'Beta' plugins or glitches to worry about.

The 'best' workflow here is obviously still a moving target, and with RED working as hard as they do, my opinion could change next week with the next update.

- Jordan

 

[Graeme Nattress]

Actually, for many reasons, you want to sharpen the image, not the raw. Sharpening in raw is not the best way to do things. I'm not at my mac with Color, so can't dig into this yet...

Graeme

 

[Jordan Livingston]

What is up for testing is A/B on signal chain.

The debayer is the same, as it is REDs.
The problem with Color has been more of a signal chain thing issue of wether the grade is basically done through adressing the metadata with extensive controll, or if the signal is locked to RGB out of the RED room, AND whether there is a full debayer or a half happening when going to 2k/1080.

A/B testing time...

But itlooks better than before...

I don't have a direct answer for this... but we have done extensive A/B testing at DIGILOID®, because for a while there, of our projects wanted to grade in RAW. After a few DPs were less than satisfied with the sharpness, especially when images were being projected on larger screens, we concluded that there was no contest: REDCine-X transcodes looked better than Color's.

This A/B testing held true for us in 1080p and in 2K. We tried it with 16:9, 1.85:1, 2.0:1, and 2.35:1 ratios. Again, it held true for all. Unfortunately, we did not test in full 4K because we did not have a display that could accurately display 4K.

But please don't take my word for it... Do your own A/Bs! I'm very interested in hearing about the results, because I'm quite excited for Color to support the REDRocket (bound to happen sooner or later!), and would love to use Color to grade RAW, all things being equal.

- Jordan