Skin tones very much off for Nikon Z8

When I compare the color rendering between the original Nikon Z8 NEF file viewed with Xnview (embedded jpg preview) and the DXO PL 7 Elite with new default profile and Nikon Z8 rendering selected, the original jpg skin tones look amazing, while DXO skin tones look unnatural, displeasing, very much on the greenish side.

I just upgraded to DXO PL7 and nearly immediately got buyers remorse, seeing how strange the portrait pictures look. Not to say that DXO PL5 was different. It is not only skin tones, but also hair brightness, which comes out too bright comparing to real life, shifting actual color of the hair.

I compared the color rendering with other software as well - Capture One 23, Nikon editing app. DXO produces the results that are most off, comparing to the original jpg previews.

I am not sure what I can do to make DXO usable. Please help.

I noticed the same with my Nikon files. For studio shots, I created my own color profile with a color-checker and also use a gray card for white balance. The colors are so much better. Am not quite sure if it is only the profiles or also the white balance that are off.

For other pictures, I noticed I use Photolab less and less and just use Nikon’s NX Studio, which basically produces the jpg output.

Have you tried with files from other Nikon camera bodies (Z6, Z7, Z9) to understand if it’s only the Z8 rendering that is tuned this way, or if it’s general to all Nikon cameras?

And viceversa, have you tried DxO’s own presets dedicated to portraits instead of the default one?

I had Z6 II before and I observed similar issues (cannot tell how much off it was to compare). General color rendering was off for people portraits.
I tried DXO portrait rendering option with Z8 in PL7 and it was less green, but still far from the original camera rendering.

I cannot tell if the problem is only with Nikon cameras or other brands as well, because I only own Nikon.

I would assume that a default DXO profile that is applied to all pictures by the software, should produce results that are acceptable for all types of pictures. Maybe I am not a sufficient expert in PL7 yet…

In my case it is the D750 and the Z6.

Ok, thanks for clarifying!

I have a Z6, and I absolutely have no issues with skin tones.

But, that must be said, I always rely on manual setting of the white balance and hue (the green/magenta slider) and never on “automatic” or “as shot”. I think this is the key for good results, and not only for portraits. And it eliminates at the root A LOT of issues with shot to shot colour variability for images taken in the same session, too.

How do you do the manual setting for the white balance and hue? I mean, how do you know what values to set for any given image, and how do you apply the values to multiple images, if they are the same values?
By comparison, in Capture One, with white balance set in camera to A0, I don’t really need to touch white balance at all. If I set it to A1, I can just click on a grey area on the image to get good white balance. I prefer A0 these days to reduce amount of work needed. Do I need to always set the white balance in PL7 one by one? Why does it not use the data from the images?

You can select multiple images from the thumbnails at the bottom of the screen, and apply the same white balance setting for all of them with a single action.

The values: for portraits either I use a grey card at the beginning of the shooting session, or I use some other grey-colored item in the scene. The grey card is more reliable and repeatable, and that’s the “pro” method.

For landscapes or other types of photos where I don’t care for realistic colour but for the most pleasing rendition, I experiment with various values of color temperature and hue until I find one that satisfies me, and then I apply these values to all photos by selecting all the thumbnails.

It does. But the automatic interpretation from an image is at the very best a “guesstimation”, and very influenced by the colors of the subject.

Unless you have measured the exact light temperature with a colorimeter, or used a grey card, it will always be “guessed” either by the camera or by the software. And each camera and each software has its own interpretation: but none is 100% precise as a colorimeter or the grey card. And, most of all, it will strongly differ between in camera JPEGs and the results produced by a third party software.

What I really don’t get is that:

  1. Camera does excellent job at getting the nice white balance and beautiful skin tones and it writes the information into the embedded JPG inside the RAW file.
  2. DXO PL seemingly ignores this information and gives subpar results in color rendering using default settings.

Have I missed something in my oversimplified analysis?
All I want is beautiful portrait pictures “out of the box” in DXO PL.

DxO and other third party NEF developers do not “ignore” that info.

Quite simply, many years ago, Nikon decided to make some of that info proprietary and “crypted”. And to make it deeply intertwined with their own Picture Controls (so it’s not only a matter of white balance and hue, but also of the very specific non linear tonal curves and saturation settings which are heavily embedded in each Picture Control).

So if you want the same exact rendition of the out of camera JPEGs (or of the embedded JPEGs) with no manual effort, there is only ONE solution: using Nikon’s NX Studio for opening the NEFs. NX Studio has access to all the hidden data tags, and to the Picture Controls, and it’s the only program that will produce the same results as the out of camera JPEGs.

No third party software will ever be able to fully replicate the same color rendition and tonal curves of the out of camera JPEGs. Not, at least, without some manual work from the user. Certainly not automatically with the default settings.

3 Likes

I have been a user of NX2/Studio for many years for the same very reasons you mentioned. Then it started to be very limiting (it lacks many features), and I moved to PL. The initial learning curve was hard, but ultimately it paid off.

It’s entirely possible that the default color rendering or white balance used by DxO needs improvement. I suggest submitting a support request at support.dxo.com.

I understand and accept your answer above.
I will need to rethink my workflow. I am not getting nice enough output from DXO PL. Maybe I will invest more time into learning how to do achieve the skin tones equivalent to embedded JPGs, or I will simply not bother with PL for portraits.

Have you tried the Portrait presets in PL instead of the default generic one?

I also read that Film Pack 7 adds some further presets to exactly emulate the Nikon Picture Controls. I haven’t tried them yet, but I would trust them to be decent (considering the quality of their other film/sensor emulations).

1 Like

Since I am already on the rant here, a while ago I contacted DXO support about an issue I observed. It was a case where enabling vignette correction would shift color of the hair - from very deep dark brown to light brown. In my world, increasing brightness of the edges of the frame should not significantly shift the color of the hair, which was at least 20% from the edge of the frame. I never got any resolution to my support case.

And just to make another example…

Saturated greens and blues are very important for me, because I have a weirdly saturated style of mountain photography.

When I moved to PL, I initially could not replicate them as I was used to. Now I have learnt how, and when I publish a printed book with mixed old and new photos, they all have those same very recognizable blues and greens.

One year ago, I was tempted to move to Capture One, and I downloaded the free trial. Once again, the greens and blues were completely off compared to what I had learnt to get from NX or PL. I decided that I didn’t want to relearn my workflow for a third time, and that was it. But this is just to say that each program has its own “flavour” of colors with the default settings.

1 Like

That makes sense. I just wish it wasn’t the case.

As a software engineer myself, an approach that immediately comes to my mind to help the software to produce “correct” skin tones would be to do one or more from the below:

  1. Compare colors between embedded jpgs to help RAW processor to generate colors
  2. Use face detection algorithms and treat skin tones in a special way, ensuring they are skin tones, not grass tones :slight_smile:

Yes, a complete overhaul to introduce AI features to analyze the content of an image and propose an automatic first attempt would be great!

1 Like