DSCF2690 (1).RAF.xmp (24,1 KB)
or with a little bit more drama:
DSCF2690 (1).RAF.xmp (30,0 KB)
5 Likes
Going for a bright rendering 
DSCF2690.RAF.xmp (37.1 KB)
PS: when I compress the history in darktable it reorders my modules I don’t know why…
11 Likes
Fantastic edit!
I tried it with some brightness as well:
But I think your render has more soul:
DSCF2690 (1).RAF.xmp (31,0 KB)
4 Likes
5 Likes
3 Likes
Which prompted the thought of Contrast-Limited-Histogram-Equalization and yet another one-click offering:
The magenta came courtesy of RT’s raw conversion. couldn’t be bothered to mess with it, sorry. Turns out that RT 5.8 doesn’t like RAF files, grump.
EDIT: I messed with it in the GIMP - Free Select then Edit>Fill with white:
1 Like
Absolutely not like in real world, but fantastic!
1 Like
You need some highlight reconstruction.
1 Like
Why? The three other apps (non-FOSS) on my computer showed no magenta at all. In any case, I’m not a great believer in highlight “reconstruction” especially when all three raw channels are blown … according to RawDigger which I shouldn’t really mention in Play_Raw.
So you don’t get that terrible magenta blob. After white balancing, when highlight recovery is not turned on, clipped highlights skew magenta. It happens in darktable too.
Those programs are likely just doing it for you without telling you. Both RawTherapee and darktable give you control over that.
You don’t have to try and recover anything, it’ll just get rid of the magenta.
1 Like
11 Likes
That first one is quite perfect imho. Your cloud recovery game is once again on point!
I guess there is an aspect of sour grapes in my little ranty screed above, now I feel like I soapboxed myself into a corner 
I wish I could get RT to do anything like that with bright clouds.
2 Likes
My first edit above I just grabbed some adjacent clouds to see what I thought…likely could do a better at selecting and cloning but it could be a way to go… For a second round I was messing with trying to keep nice fog and let the sky blow out without that small obvious patch I’m not sure I like these edits either but I went with the let it blow out sort of… I think many of the ones trying to mask it with some yellow really don’t work for my taste and might actually draw more attention than just blowing it out… I think maybe the secret for a believable look in this one is just getting it to nicely blend in / fade in the blown area into the surround sky… Mine would need more work there for sure…
DSCF2690.RAF.xmp (17.3 KB)
DSCF2690_01.RAF.xmp (18.8 KB)
2 Likes
I’m quite sure, it is possible with RT as well. In the past RT was better than dt regarding hihglight reconstruction: Just have a look here: Highlight reconstruction darktable vs. rawtherapee
I bet, the RT developers have as well improved it a little bit over the time so that both programs should be on par nowadays.
1 Like
nice capture, here my version with RT dev, by trying to emphazise the rays under the clouds:
DSCF2690.RAF.jpg.out.pp3 (17,5 KB)
btw: happy new year 
3 Likes
7 Likes
So… what you’re saying is that I’m simply too bad at using RT. Got it, thanks! 
I hope that it is like that. Because then you just have to train a bit, 
But it is too lang ago when I last time used RT for such tasks. So I can’t really tell you.
I’ll give it a try the next days. Even so I fear my capabilities with RT are far away from what they were in the past.
But looking at your edits, I think you are complaining on a very high level. There is not much wrong with highlight reconstruction. The only thing I would change is to give the highlights a bit of colour.
1 Like
5 Likes
The magenta comes from:
- in raw, all channels in an area are blown, so all pixels have value = 1
- then comes a white balance adjustment with RGB multipliers 1.917, 1, 1.517 → the blue pixels in the blown area will be pushed to 1.917, the green remain 1, the red are pushed to 1.517
- once you demosaic that, you get (1.917, 1, 1.517) → purple
- software that simply clips values to 0…1 will show it as white (1, 1, 1)
- software that can handle unbounded values (such as RawTherapee, ART and darktable) would show purple
- proper highlight reconstruction can turn that neutral (e.g. rendering it as 1.7, 1.7, 1.7), where it can still be differentiated from diffuse white (1, 1, 1) . You won’t recover the colour, but you will recover detail in the sense that it will be brighter than non-clipped neutral pixels (it’s OK to have the Sun or a lamp being brighter than the brightest diffuse white, like snow or a white piece of paper or hair/fur etc).
Or suppose you have a bright blue area. Let’s say that in the raw, the blue pixels are close to clipping, at 0.9. Green pixels are at 0.5, red pixels at 0.33.
After WB + demosaicing you get 1.73, 0.50, 0.50.
Software that clips to 0…1 will change that to 1, 0.5, 0.5 - you lose information and distort colours. Software handling unlimited values will just carry on.
With software using values bounded by 0…1, you can then reduce exposure, darkening the image until no non-clipped raw value produces a value above 1 after WB + demosaic, and apply curves to brighten the shadows, being careful not to push anything above 1, ever. That’s basically display-referred editing.
3 Likes