R5 colour science and skin tones.

Feb 15, 2020
324
238
After watching quite a bit of video footage and seeing images shot with the R5, I've noticed a difference in the colour science and skin tone rendering when compared to the EOS R or 5D Mark IV. I have always loved the colours I've gotten out of the EOS R and 5D Mark IV, so it was a little disheartening to see Canon decided to make changes to the R5's colour science. Skin tones and neutrals now seem to have a magenta cast.. similar to what the 5D Mark II and Mark III had.

If anyone who owns the camera has some more examples of skin tone rendering with the R5 I would love to see them!
 

StoicalEtcher

EOS RP
CR Pro
Jan 3, 2018
331
266
Yorkshire
After watching quite a bit of video footage and seeing images shot with the R5, I've noticed a difference in the colour science and skin tone rendering when compared to the EOS R or 5D Mark IV. I have always loved the colours I've gotten out of the EOS R and 5D Mark IV, so it was a little disheartening to see Canon decided to make changes to the R5's colour science. Skin tones and neutrals now seem to have a magenta cast.. similar to what the 5D Mark II and Mark III had.

If anyone who owns the camera has some more examples of skin tone rendering with the R5 I would love to see them!
Bear in mind Chris that I'm not sure the software is properly there yet to deal with R5 files is it? Don't have one myself yet, but I didn't think Adobe CS was ready yet (or maybe only just now) - so I wonder how much rendering is still being done using proxies from other cameras (such as 5Div) which may not be spot on?

May be worth asking if anyone has yet been processing files using whatever is your go-to software, and see how those compare, if yet available?

Cheers. Stoical.
 
  • Like
Reactions: Chris.Chapterten

Viggo

EOS R5
Dec 13, 2010
4,569
1,261
I use a ColorChecker always and make my own profiles to match, so it’s not always that important . However, the more the profile changes the raw the more noise you get, so it will be interesting to check out when I eventually get my hands on an R5.
 
Feb 15, 2020
324
238
Bear in mind Chris that I'm not sure the software is properly there yet to deal with R5 files is it? Don't have one myself yet, but I didn't think Adobe CS was ready yet (or maybe only just now) - so I wonder how much rendering is still being done using proxies from other cameras (such as 5Div) which may not be spot on?

May be worth asking if anyone has yet been processing files using whatever is your go-to software, and see how those compare, if yet available?

Cheers. Stoical.
I think that's a fair point for the RAW files of photos. I'll have to wait and see how they look once adobe camera raw fully supports the files. However the thing that has me a little concerned is the magenta cast I'm seeing in the video files (which has nothing to do with RAW converters etc.).

Very curious what R5 owners think about the skin tones in video too. Cheers
 
  • Like
Reactions: StoicalEtcher

analoggrotto

EOS RP
Aug 27, 2016
228
125
Interesting, I never noticed that personally, but I also didn"t shoot many red subjects. What did you usually have to do to correct them?
I usually start with temperature then the R-G-B calibrations at the very bottom of the lightroom panel. I iterate next to the individual saturations. I also have never read any tutorials or watched any youtube videos (unless pewdiepie would do one lol), just kept at it by myself until the results were satisfying. And they are.

X18_7961-Pano-Edit.jpg
 

Act444

EOS R
May 4, 2011
1,115
189
I don't like the way the 5D4 renders reds either. the 5DSR is MUCH better in this regard (as was the 5D3). Unfortunately, based on the DPReview tests, the R5 appears to be even worse.

(If anyone knows any color profiles in DPP that allow 5D4 or Rx files to mimic 5D3 colors, I'm all ears)
 
  • Like
Reactions: Chris.Chapterten

analoggrotto

EOS RP
Aug 27, 2016
228
125
The 5DS/R was the last of Canon's non Dual Pixel, offboard ADC, sensors. Much changed after that.

If I ever get my hands on an R5, I'd love to put it to the test.

Also, I tried shooting in AdobeRGB versus sRGB and it just did not work out.
 
  • Like
Reactions: Chris.Chapterten

Act444

EOS R
May 4, 2011
1,115
189
I think the change in color science may be somehow tied to the Dual Pixel technology. Cameras with DP have the duller, flatter native color output (compared with previous generations). This started with the 5D4/M5/M6 generation. Now, with DP II hitting the scene with the R5/R6 generation of cameras, I'm seeing a further erosion in color reproduction of yellows and greens (and to a lesser extent, reds). Still much better (IMO) than the yellowish cast present on Sony's cameras, but a bit disappointing nonetheless to step backwards. I probably wouldn't even bother to mention it if it were easy to replicate the old colors in post, but it has proven to be a bit tricky, as a poster above pointed out.

Think that's what helps make the 5DS/R my favorite portrait-type camera thus far (apart from the resolution); the RAW files need only minimal tweaking (in most cases), very sharp and the strong OOC reds and yellows make for nice warm skin tones.
 
  • Like
Reactions: Chris.Chapterten

privatebydesign

Garfield is back...
CR Pro
Jan 29, 2011
9,036
3,109
120
Always felt reds were difficult on the 5D Mark IV. Really had to spend time in post to get them right.
Interesting, I never noticed that personally, but I also didn"t shoot many red subjects. What did you usually have to do to correct them?

I usually start with temperature then the R-G-B calibrations at the very bottom of the lightroom panel. I iterate next to the individual saturations. I also have never read any tutorials or watched any youtube videos (unless pewdiepie would do one lol), just kept at it by myself until the results were satisfying. And they are.

View attachment 191978
Technically reds can be difficult on all regular Bayer array sensors because the red channel normally clips first which means you get blocks of red rather than subtle tonal transitions, just watch your RGB histogram for clipped red channel issues. The hue is easily and very quickly dealt with by using a ColorChecker passport, no guesswork, no wasted time, just accurate color.

The car above is a perfect example of clipped red channel blocking, open it in PS or LR and run the dropper tool over it and see how much of it is red channel 255, way too much.
 
  • Like
Reactions: SecureGSM

privatebydesign

Garfield is back...
CR Pro
Jan 29, 2011
9,036
3,109
120
The 5DS/R was the last of Canon's non Dual Pixel, offboard ADC, sensors. Much changed after that.

If I ever get my hands on an R5, I'd love to put it to the test.

Also, I tried shooting in AdobeRGB versus sRGB and it just did not work out.
If you shoot RAW and use Lightroom capturing in AdobeRGB or sRGB makes no difference to the image, Lightroom uses it's own colorspace in which to process RAW files. Having said that the choice will impact the in camera preview and the in camera histogram. If you want to use your channel histogram to warn about channel clipping then set the camera to AdobeRGB as it will give you a better and more accurate indication than sRGB will.
 
Last edited:

analoggrotto

EOS RP
Aug 27, 2016
228
125
Technically reds can be difficult on all regular Bayer array sensors because the red channel normally clips first which means you get blocks of red rather than subtle tonal transitions, just watch your RGB histogram for clipped red channel issues. The hue is easily and very quickly dealt with by using a ColorChecker passport, no guesswork, no wasted time, just accurate color.

The car above is a perfect example of clipped red channel blocking, open it in PS or LR and run the dropper tool over it and see how much of it is red channel 255, way too much.
If it is too much, it is deliberate. I wanted to match my memory of how I thought I saw the car; which of course is subjective. But I'll give your advice a shot.
 
  • Like
Reactions: stevelee

privatebydesign

Garfield is back...
CR Pro
Jan 29, 2011
9,036
3,109
120
If it is too much, it is deliberate. I wanted to match my memory of how I thought I saw the car; which of course is subjective. But I'll give your advice a shot.
I'm certainly not criticizing, merely trying to point out a different way to process so there is no need to "Really had to spend time in post to get them right." If we can't help each other out with stuff other than how fast the R5 overheats then the forum will be a very negative place! :)

The Porsche in your picture is Guards Red, here is the official picture from Porsche of a Guards Red 911.
Screen Shot 2020-08-09 at 11.33.35.png


If you put it into Photoshop and make a colored fill layer on top with red at 255 and set the blend mode to 'difference', in an optimally exposed red channel image you will see a little black but mostly gradations, you don't want blocks of black.

It will look like this. Small black patches on the tops of the fenders hood, C-pillar etc but quickly fading to nice gradations.
1596987955839.png


With the same fill layer your image looks like this. Large black patches on the three fenders, the hood the roof and the front bumper.
1596988193001.png


To be clear, there are two different aspects at play, one is WB which as you say is personal to the day and specific light source, that accounts for the differing hue, but the blocking (lack of tonality) is due to a clipped red channel which can only be fixed by less exposure at capture it can't be fixed in post.

Sorry if I have made a pain of myself, truthfully I consider this a learning site and with there having been so much negativity here recently I thought it appropriate to try to give out some positive comments. Reds are defiantly an issue for many digital cameras and knowing how to get them right if and when we need to can make a huge difference and save a ton of time.
 

SteveC

R5
CR Pro
Sep 3, 2019
1,599
1,449
I'm certainly not criticizing, merely trying to point out a different way to process so there is no need to "Really had to spend time in post to get them right." If we can't help each other out with stuff other than how fast the R5 overheats then the forum will be a very negative place! :)

The Porsche in your picture is Guards Red, here is the official picture from Porsche of a Guards Red 911.
View attachment 192024

If you put it into Photoshop and make a colored fill layer on top with red at 255 and set the blend mode to 'difference', in an optimally exposed red channel image you will see a little black but mostly gradations, you don't want blocks of black.

It will look like this. Small black patches on the tops of the fenders hood, C-pillar etc but quickly fading to nice gradations.
View attachment 192025

With the same fill layer your image looks like this. Large black patches on the three fenders, the hood the roof and the front bumper.
View attachment 192026

To be clear, there are two different aspects at play, one is WB which as you say is personal to the day and specific light source, that accounts for the differing hue, but the blocking (lack of tonality) is due to a clipped red channel which can only be fixed by less exposure at capture it can't be fixed in post.

Sorry if I have made a pain of myself, truthfully I consider this a learning site and with there having been so much negativity here recently I thought it appropriate to try to give out some positive comments. Reds are defiantly an issue for many digital cameras and knowing how to get them right if and when we need to can make a huge difference and save a ton of time.
I found this educational, but of course the person whose work is being critiqued can oft-times be uncomfortable with the honor. I just try to remember when I am the one who is on the hot seat that at least my effort was worth trying to improve on, rather than hopeless. I've been the critiquer in some endeavors a number of times, and I hate when something is so bad you don't know where to begin. (I am reminded of a specification I wrote at work...it went to someone for review and came back with nearly 200 annotations--good ones. When I went to go see the guy to ask a question about one of them, he actually mimed patting me on the head--in spite of the two hundred comments he thought I had done a good job.
 

analoggrotto

EOS RP
Aug 27, 2016
228
125
Sorry if I have made a pain of myself, truthfully I consider this a learning site and with there having been so much negativity here recently I thought it appropriate to try to give out some positive comments. Reds are defiantly an issue for many digital cameras and knowing how to get them right if and when we need to can make a huge difference and save a ton of time.
Absolutely not, I too found your explanation and exhibits to be helpful. I agree about the clipping, and probably pursue some further investigation when I find the RAW file for that picture. Thanks for taking the time for that.

When I get my R5, if it does not light ablaze and kill the mail carrier, sending me in federal penitentiary, I'll start out with it set to AdobeRGB.

Cheers

(edit, minor typos and w/c)
 
Last edited:
Feb 15, 2020
324
238
It looks like a big reason for the red clipping in the Porsche photo is because it has been shot in direct sunlight. The official porsche picture is under studio lights which would no doubt be much softer than direct sun and would have allowed for those gradations in the paint.

After receiving my R5 and taking a few portraits, I am less concerned abou the magenta cast I had noticed previously. After converting the images with the Adobe DNG converter and using the "adobe color" profile, moving the green/magenta slider to 0 or +5 seems to have removed the colour cast completely (default starting point was +14 magenta).

I'll be doing some more testing in mixed lighting conditions to get a better feel for the camera... will let you know how I get on.
 
  • Like
Reactions: analoggrotto

privatebydesign

Garfield is back...
CR Pro
Jan 29, 2011
9,036
3,109
120
It looks like a big reason for the red clipping in the Porsche photo is because it has been shot in direct sunlight. The official porsche picture is under studio lights which would no doubt be much softer than direct sun and would have allowed for those gradations in the paint.
No, the light source is irrelevant the exposure is the important part whatever the lighting and for a very saturated color it is critical to look at the channel histogram to get the optimal exposure. The studio shot will have had complete control over the lights that is true, but the blended layer illustrates that they created that light and subsequent exposure optimally for their red channel.

Put another way, in the studio they had complete control over both the lights and exposure, for the sunlight shot you can only control the exposure, ergo any problem is an exposure problem.
 
Feb 15, 2020
324
238
No, the light source is irrelevant the exposure is the important part whatever the lighting and for a very saturated color it is critical to look at the channel histogram to get the optimal exposure. The studio shot will have had complete control over the lights that is true, but the blended layer illustrates that they created that light and subsequent exposure optimally for their red channel.

Put another way, in the studio they had complete control over both the lights and exposure, for the sunlight shot you can only control the exposure, ergo any problem is an exposure problem.
That's true..

Another thing to keep in mind with the official porsche image is that a lot of the gradients we are seeing are probably created in post. Very common to do so in Automotive photography. Unwanted reflections and highlights are often cloned out.
 
Feb 15, 2020
324
238
Well I have tested the R5 in some different lighting conditions while taking portraits. Initially I felt the colours were quite neutral but upon further testing there is definitely a magenta colour cast. This colour cast is present to the same extent whether using the Adobe DNG converter (set to 'Adobe Color' in camera raw) or Canon DPP4 (set to 'camera neutral')

I tested using natural window light and a Litepanels Gemimi 2x1 LED panel set to 4000k (a setting in which the light is completely neutral in terms of green / magenta shift). I found with the DNG file through Adobe DNG converter I was having to dial in -15 magenta (+15 green) on the slider to neutralize the colour cast. In DPP4 I was having to dial in +5 green (out of a maximum 10).

Overall I would say the magenta colour cast is less than what is present on the 5D Mark II but about the same as the 5D Mark III.

Honestly I'm a little dissapointed with this as I really loved the colour rendition on my 5D IV and EOS R, I found them to be very neutral for portraits. I will keep doing some further testing.. but for now I have the R5 set to +5 green in the cameras white balance settings to account for the magenta shift and it seems to be doing the job for me. My only concern is that if the Magenta cast is mostly affecting skin tones and not background / foreground elements then I may introduce a green colour cast to other areas of the image as a consequence of correcting for the skin.

I'll report back once I've had a chance to test the R5 on a fashion shoot under even more varied lighting conditions.