RawTherapee 5.9 apparently calculates white balance slightly differently in comparison with the previous version 5.8. For my RAW files and others I observed that I need to subtract 0.04 from the Tint slider in the White balance module in order to obtain a visual match with previously exported files. That’s simply “wonderful” – in the worst case I need to perform the correction on several hundred important for me photos. These are not family photos nor pet photos.
I am not aware of any white balance calculation changes between 5.8 and 5.9, except for the addition of a new automatic method. Then again, the time has been long and changes have been many.
Is there an example file and pp3 sidecar that you can share from 5.8 that looks unexpectedly different in 5.9?
I noticed something similar. I believe the difference may be down to the addition of the ‘automatic method’ that @Thanatomanic has mentioned. In my case, though, the difference was very welcome; I found the results to be a significant improvement.
The tint shift is small, but perceptible. One example is my Play Raw edit - Handling shade + direct sunlight with skin-tones, filmic / sigmoid preference? - #22 by teplit
You can substitute the likely missing AdobeRGB1998 output profile with RTv4_Medium - the result is the same. Since the difference isn’t big, compare the exported JPG images one on top of the other. Take a look at the woman face or the window frame behind her.
It’s not OK to alter the image rendering in the newer version without notice.
I can confirm the color shift. Let me try to see what the cause of this is.
I have just filed an issue for this on GitHub: Regression? Color shift even though automatic wb is not used. · Issue #6639 · Beep6581/RawTherapee · GitHub
Let’s see if we can find out what changed, and why.
Hi,
Unless something has changed recently, the wb channel multipliers calculated by RT depend on the standard camera matrix used. So, if you updated the camera matrix, old values of temp/tint will result in different coefficients. I don’t know if this is what is happening here but I thought it worth mentioning…
HTH
If it only were that simple 
The commit I traced it to seems to be related to automatic white balancing, which the example doesn’t use…
Got it – I’ll learn to shut up
I made a little experiment with GFX100 sample. I don’t have medium format camera, nor work with such files. My main camera is Samsung NX500.
With RawTherapee 5.8 I opened the sample RAF file and set the Temperature to 7200. Then I exported the following image:
Samuel_Zeller_DSCF3239 (RawTherapee 5.8).jpg.out.pp3 (12.1 KB)
After that I opened the same RAW file with RawTherapee 5.9 and without making any change I exported the following image:
Samuel_Zeller_DSCF3239 (RawTherapee 5.9).jpg.out.pp3 (14.3 KB)
With pink tint the image is more “nice”, don’t you think so?
@teplit Thanks for reporting! It seems there really was an oversight that nobody ever detected in their output. @jdc has provided a quick fix. It will certainly be included in the next minor release of RT and will likely be available soon in an automated build.
3 Likes
I would to thank @Thanatomanic and @jdc for the rapid response to this issue. Although sometimes I express criticism of RawTherapee, I find it for valuable tool (it is my main RAW software). The new automatic white balancing method for example is in my opinion very good improvement.
About the Fujifilm GFX 100 sample - initially I thought that the pink tint problem is somehow related to the discussed issue, but it appears that the problem is different… Anyway I decided to leave it in case someone is interested to investigate this problem (I’m not interested).
2 Likes
This is a separate issue indeed. For some reason, the color matrix is weird (it is caused by camconst.json : added some dcraw_matrices from ART · Beep6581/RawTherapee@6490c7d · GitHub). I’ll take a look again.
Actually, I think this was a problem in 5.8 and is now correctly done in 5.9. In your pp3, the tint value is set to ~0.5, which is extremely low. This may have been necessary back then to compensate for the lack of proper color matrix. Now that we have that (since 10 March 2020 in dev), the low tint causes a strong magenta cast.
1 Like
I was just going to post an identical hypothesis; it would certainly explain my personal experience.
I compiled a development version of RawTherapee and tried it with RAW files from my cameras. The issue is resolved - the greenish tint is gone.
I think that part of the problem here is the way RawTherapee works. By design Camera Standard as Input Profile can mean at least three different things and it is not clear which one of them is in effect. One way to alleviate the problem is to program RawTherapee to warn the user that the input color matrix is effectively temporary (subject to change in future versions). A strange white balance in my opinion is neither a reliable nor user-friendly way to warn the user about potential compatibility issue in the future.
1 Like