How to adjust lightness of red channel only?

Im working with a GH3 camera, and the reds are just a little too bright and hot. This makes skin look slightly unnatural and flat. I want to find a way to isolate the highlights and midtones of the red channel only, not just the red pixels in the clip, and then just darken them slightly.

What is the best way to do this? I have tried setting the corrector node to RGB color space and channel 1 only, and then tried to do some adjustments, but it seems to be shifting the color instead of just darkening it.

Anyone have any ideas?

Use the sliders in the channel mixer. (the one with the tree circles left of camera raw)

But I'll hijack the thread for a simlar question.
I read a tutorial about simulaton of three strip technicolor. The tutorial sucks for two reasons (first the look cannot be applied by simple linear transformation and the guys didnt realize that subtractive "addition" is multiplication), but I realized a problem.

It seem no possible, due to the YRGB to make pure RGB calcualtions and adjustments like in Photoshop.
Say. When I have a black and white input image and I turn the gradation curve of green and blue to maximum (horizontal) an keep the red curve, I expect something like cyan where the b/w is black and white where the input b/w is white. But it isn't. It contains additional luminance infos due to the Y-Channel.

YRGB is good for grading, but not for calculation like this.

Thomas

Use the sliders in the channel mixer. (the one with the three circles left of camera raw)

But I'll hijack the thread for a simlar question.
I read a tutorial about simulaton of three strip technicolor. The tutorial sucks for two reasons (first the look cannot be applied by simple linear transformation and the guys didnt realize that subtractive "addition" is multiplication), but I realized a problem.

It seem no possible, due to the YRGB to make pure RGB calcualtions and adjustments like in Photoshop.
Say. When I have a black and white input image and I turn the gradation curve of green and blue to maximum (horizontal) an keep the red curve, I expect something like cyan where the b/w is black and white where the input b/w is white. But it isn't. It contains additional luminance infos due to the Y-Channel. The Cyan blackend.

YRGB is good for grading, but not for calculation like this.

Thomas

I don't think the channel mixer will work. I want do darken the red channel, not change its mix level...

EDIT: After looking at it more, I think the only way to achieve this is with the HUE vs LUM mixer.

I don t understand you. Coming from (linear) color correction of photoshop, darken the red channel is identical to make the image (darker and) more cyan.

The YRGB model is different, thought.
So what exactly do you mean.

(you noticed that in the channel mixer is a flag "preserve luminance" or so.
With this switch on (default) the behavior is different. Darken the red channel will only make it cyan without darkening.

Use the sliders in the channel mixer. (the one with the three circles left of camera raw)

But I'll hijack the thread for a simlar question.
I read a tutorial about simulaton of three strip technicolor. The tutorial sucks for two reasons (first the look cannot be applied by simple linear transformation and the guys didnt realize that subtractive "addition" is multiplication), but I realized a problem.

It seem no possible, due to the YRGB to make pure RGB calcualtions and adjustments like in Photoshop.
Say. When I have a black and white input image and I turn the gradation curve of green and blue to maximum (horizontal) an keep the red curve, I expect something like cyan where the b/w is black and white where the input b/w is white. But it isn't. It contains additional luminance infos due to the Y-Channel. The Cyan blackend.

YRGB is good for grading, but not for calculation like this.

Thomas

If you set the lum mix to 0 you have a RGB math and not the YRGB math

Im working with a GH3 camera, and the reds are just a little too bright and hot. This makes skin look slightly unnatural and flat. I want to find a way to isolate the highlights and midtones of the red channel only, not just the red pixels in the clip, and then just darken them slightly.

What is the best way to do this? I have tried setting the corrector node to RGB color space and channel 1 only, and then tried to do some adjustments, but it seems to be shifting the color instead of just darkening it.

you are literally describing the action of putting a point in the red main curve about in the middle and pulling it down... That will reduce the "lightness" without change hue. This will make the reds "darker" and more "fuller".

Another way (as someone was describing) is to use the Hue vs Luma curve.

If you set the lum mix to 0 you have a RGB math and not the YRGB math

Thanks!

I have overlooked this.
Works great. Good to know.

If you set the lum mix to 0 you have a RGB math and not the YRGB math

The manual indicates that if you set lum mix to 0, changing one unganged curve does not effect the others, but that's not what happens on my system. Any change in R, G or B curves effects the others regardless of the lum mix setting.

The only thing lum mix seems to do is modify the degree to which the Y only setting (lum life, lum gamma, lum gain) controls the image as compared with the YRGB setting.

If you understand the math difference between RGB and YRGB, an explanation would be much appreciated.

Cheers.

If you set the lum mix to 0 you have a RGB math and not the YRGB math

The manual indicates that if you set lum mix to 0, changing one unganged curve does not effect the others, but that's not what happens on my system. Any change in R, G or B curves effects the others regardless of the lum mix setting.

The only thing lum mix seems to do is modify the degree to which the Y only setting (lum life, lum gamma, lum gain) controls the image as compared with the YRGB setting.

If you understand the math difference between RGB and YRGB, an explanation would be much appreciated.

Cheers.

... if you have a "yellow" color and you change just the red, you will change the relationship between the red you just changed and the green component used to make that yellow, regardless of YRGB/RGB situation.

Does make sense?

This whole approach smacks of a Photoshop-locked mindset and is not generally how motion picture grading is done in DaVinci. Individual channel grading is something that was only introduced with Custom Curves sometime in the 8:8:8 era, and it is still of somewhat limited value in a highly-time dependent workflow. I got the most use out of unlocking the channel curves only really when dealing with some seriously-damaged negative stocks or massively faded (ie. cyan layer gone) archive prints.

Perhaps my own paradigm is colored by that inertia/momentum, but the desired action, at least as I understand the language used to describe the desired end result, would indicate a secondary isolation and at that point, a primary correction within that qualification would be the first solution an experienced motion picture colorist would likely implement. First, it unambiguously segregates the area you want to affect and then it gives the colorist total, subtle control over the target values while minimizing the likely deleterious noise artifacts that curve adjustments can sometimes introduce. daVinci (small d) invented Kilovectors and Power Windows to do this and fast, because we needed the corrector to be able to do it, and on the fly. Custom Curves were for creating filmic S-curves to modify the gamma characteristic that we usually got stuck with.

This is also the general approach to how I fix blown-out faces, so that they don't look like the actors have had the skin peeled off their skulls, where there is either no color because the values are in the clipper or entirely the wrong value because the camera has failed into a semi-random hue that is a combination of highlights that are semi-clipped or worse, a cross-light of a different key exposure or temperature. Which is very very often. To be honest, I do use the hue/hue and hue/sat curves (analog to the old Kilovectors) but make sure they are on a separate, following node -- I do see unwanted side effects within corrections if you try too many adjustment types within a node. It just makes more sense to compartmentalize adjustments anyway so that you can copy and paste them, de-select them... click them on and off to before/after, change the concatenation... but thats maybe just a slavish orthodoxy.

jPo

JPO, your way seems reasonable as approach, although not the only one, it almost seems you are talking like is the "only" way to do it...

I think that the art of a colorist is also interpretative of the desire of the client and understanding the language converting that in a real action (button presses). Bashing as "approach smacks of a Photoshop-locked mindset " does not look constructive to me. (But a usual is just an opinion)

So, if up to this point we gave more or less instruction in how to perform a task, why we don't approach in understanding what the client is trying to tell us?

and the reds are just a little too bright and hot. This makes skin look slightly unnatural and flat. I want to find a way to isolate the highlights and midtones of the red channel only, not just the red pixels in the clip, and then just darken them slightly.

Hot might indicate too saturated, the hue drifting to red, too bright respect the desired value or a combination of them all.

So, I think that the OP should post a image and we can see. Instinctively I will adjust the whole image (comtrast/tonal curve/brightness/saturation/black balance/white point) until the color is coherent with the scene, then I will see where the skin lay. Probably I will have at that point one of two approach it, either use curves to deal with the hue/luma/sat changes, or I will make an extra node, select very softly that color and using primaries grades shift it where I want it...

jjsgt84, what's your take?

Acknowledged, as written could be construed as "This is the only way it can be done", and I need to shake off my own orthodoxy from time to time.

Of course it isn't, but assuming that Resolve could or should be used in a way that Photoshop accomplishes things needs to be re-evaluated as a mind-set, as well.

The technique described is just a go-to that fits inside the daVinci architecture, does the least harm (usually) and goes up the flagpole easy and fast, because I have another 1200 events to attend to.

jPo

Of course it isn't, but assuming that Resolve could or should be used in a way that Photoshop accomplishes things needs to be re-evaluated as a mind-set, as well.

jPo

Agreed with you but I think that most artist will start with what is familiar from them and adapt to the new situation, Human nature perhaps?

JPO, your way seems reasonable as approach, although not the only one, it almost seems you are talking like is the "only" way to do it...

I think that the art of a colorist is also interpretative of the desire of the client and understanding the language converting that in a real action (button presses). Bashing as "approach smacks of a Photoshop-locked mindset " does not look constructive to me. (But a usual is just an opinion)

So, if up to this point we gave more or less instruction in how to perform a task, why we don't approach in understanding what the client is trying to tell us?

and the reds are just a little too bright and hot. This makes skin look slightly unnatural and flat. I want to find a way to isolate the highlights and midtones of the red channel only, not just the red pixels in the clip, and then just darken them slightly.

Hot might indicate too saturated, the hue drifting to red, too bright respect the desired value or a combination of them all.

So, I think that the OP should post a image and we can see. Instinctively I will adjust the whole image (comtrast/tonal curve/brightness/saturation/black balance/white point) until the color is coherent with the scene, then I will see where the skin lay. Probably I will have at that point one of two approach it, either use curves to deal with the hue/luma/sat changes, or I will make an extra node, select very softly that color and using primaries grades shift it where I want it...

jjsgt84, what's your take?

Thank you for the continued conversation, everyone. It is educational.

This thread is where this discussion of the red channel began, please look at some of the examples on this page, and others around it. It will help you get a better idea of the issue.

http://www.personal-view.com/talks/disc ... ttings/p19

Sorry but that thread is somewhat monstrous.

Making a "random" tweak lut +nCorrections to fix a problem does not seems the best approach to me, I will start form the raw file, understand which is the tonality mapping, the three primaries and make a correct color transformation from the original color space of the camera (native) to the desired display referred color space...

After that we can talk about how the camera "see the colors"..

I've not seen any footage personally from that camera, this is just one approach that I will comfortable to start with.

Im working with a GH3 camera, and the reds are just a little too bright and hot. This makes skin look slightly unnatural and flat. I want to find a way to isolate the highlights and midtones of the red channel only, not just the red pixels in the clip, and then just darken them slightly.

Alexis Von Hurkman has an excellent section in his Color Correction Handbook on how to deal with clipped reds, using a layer mixer to map the green or blue channels in to restore lost detail in the reds. Very useful idea. Pages 170-174 of the 2nd edition.