Why use exposure and not basic adjustments? I like the auto feature in basic adjustments for a ultra quick starting point, and as a bonus I get contrast, saturation and vibrance.
Generally I think a quick start guide would be helpful to people new to darktable. I often read that people get it, play with it a little, delete it.
This article tells you which modules are recommended and why and is not outdated: PIXLS.US - Darktable 3:RGB or Lab? Which Modules? Help!.
You can still use all of the modules but some must be used with care (don’t push the sliders too far) and, as with anything in darktable, there is usually more than one way to do something, so prefer the methods recommended in this article.
That is my understanding too, I may have been incorrect in saying “it doesn’t bring back detail lost through clipping exposure”, but I have certainly seen detail lost that couldn’t be recovered - it’s why I changed my workflow. Perhaps there was another module used in between exposure and filmic which clipped values irretrievably. So the main difference is that middle gray adjustment occurs late in the pipe, whereas exposure occurs early. Perhaps if you boost exposure beyond 255, and don’t use any modules in between it and filmic that irretrievably clip the data, then your process will be good. But if you push values beyond 255 then clip them in another module, they are lost to filmic. So how do you know which modules clip irretrievably and which don’t? I’m quite sure the current channel mixer does (but the new one in the works won’t). Not sure about the other modules.
Hi,
thanks for the reply. This is exactly what I do, but getting good results is not quite easy and requires a lot of going back and forth.
Hi,
thanks, I agree with your comment that mapping the black and white level to the min/max will change considerably the image look for some pictures, but I thought you could then refine the tone mapping via the Filmic module.
As for the automatic masking of the clipped-in-raw areas, for the purpose of highlight reconstruction, isn’t the degree of fading out of the mask already controllable by the Transition slider?
Thank you, it does make sense.
Thank you. When you say “In Filmic, adjust highlights and shadows”, do you mean the White Relative Exposure and the Black Relative Exposure sliders in the “scene” tab? And do you not touch the Look > Contrast slider at all?
Thank you, that’s helpful!
I think the critical point here may be to set the correct clipping mask for highlight reconstruction somewhere close to the areas clipped in raw?
This is interesting. So, if the Exposure and Middle Gray slider are in fact positioned earlier and later, respectively, in the pixel pipeline, it would make sense indeed to use the Exposure to avoid clipping and fine tune later with the re-enabled middle gray slider. Could somebody confirm that this is the case?
Great thread, I’m getting a lot out of it.
With regard to black level, I try to set this in filmic or even later in other modules and not via exposure module.
Pay no attention to clipping highlights or shadows [in the exposure module]
The manual states to avoid negative pixels before using filmic but it may be out of date or I may be misunderstanding it :
However, it is important to avoid negative pixels in black areas…brighten the blacks by setting a negative black level value in the exposure module.
https://darktable.gitlab.io/doc/en/tone_group.html#filmic
I did some testing in DT 3.2 using the same image, one to ignore negative pixels via the exposure module and the other to manage them.
Test 1 ignore negative pixels
Exposure +0.74EV, black level correction default -0.0002
In filmic set white to +3.67EV and black to -13.85EV
Test 2 manage negative pixels
Exposure +0.74EX, black level correction -0.0062
In filmic set white to +3.67EV and black to -7.75EV
To my eye I end up with identical images but test 1 just feels wrong because of the extreme black setting as it implies a dynamic range of 17.5EV, way beyond the humble Canon 100D.
This is not so much about reconstructing highlights so much as retaining detail that was already there. Again, someone will know more than me on this, but I only use highlight reconstruction when the raw sensor data is clipped. If using filmic reconstruction for that, then yes, I set the mask so it roughly matches the area of raw clipped data. If raw data isn’t clipped, then highlights should be kept under control by filmic ‘scene’ tab (and careful use of other modules), and shouldn’t need reconstructing.
The order of pixel pipe is bottom to top in ‘active modules’. My understanding is if exposure pushes values above 255, and you have a module between it and filmic that clips data irretrievably, then filmic can’t bring it back - in which case its best to avoid clipping in exposure, and fine tune with the middle grey in filmic. However, if there is no module that clips data irretrievably between exposure and filmic, then pushing above 255 in exposure is fine as filmic white relative exposure can bring it back, and it is best to do as paperdigits said - expose for the midtones without worrying about the highlights. But I could be wrong, so someone else will need to confirm that for me too!
Yes, correct.
You can add contrast that way as well, but I generally don’t need to touch that slider unless the file is extremely flat.
I am not sure on this I do know that if you enable the gray slider first and then go to the scene tab and use the auto picker…it gives a pretty good starting point and auto adjusts the grey value widely so you don’t need to go back and mess with exposure. I have had pretty good luck doing this way and then minor tweak to the white slider and middle gray to tweak the auto to taste…every now and again it give a wonky value for black but I just set it back to around a default value and then tweak a bit if needed…of course you could tweak it every time…I would say that highlight recovery…may be improved or more accurate but it does not appear as dramatic or as potent as what you can achieve in ART or RT…at least up to this point in my hands…
Yes, but I only stated that it is difficult to automatize because in requires choice.
The default for the threshold is 3EV which is essentially off…since the module is so CPU intensive so by default filmic is not really doing HR until you drop that …
Well… straight from the horse’s mouth, as they say 
Thank you very much!
Thank you all.