Lost highlights in darktable


on some old RAW images that I’m editing with darktable, I notice that the highlights are gone. When I compare them with my old edits in capture-one I see that the information is there. For example:

With capture one:

With darktable:

You can notice that close to the arrow on the ground there are some clipped highlights that are not recognized as such in capture one.

I could not manage to get filmic-rgb/reconstruct to recover properly those highlights.

Has someone any idea why this is happening?

Thank you!

The original RAW:
_DSC7880.ARW (24.2 MB)

The raw file is licensed Creative Commons, By-Attribution, Share-Alike

How about something like this?
The key is to not use highlight reconstruction with clip highlights. Set it to reconstruct in LCh or disable it altogether (but then you’ll have to deal with the magenta highlights e.g. in filmic).

Or this:
_DSC7880.ARW.xmp (7.8 KB)

_DSC7880.ARW_01.xmp (10.1 KB)


Because the reconstruct there is meant to smooth over blown patches that are causing issues, not to bring detail that is in the file back into visible range.

If the data is there you just need to play around with the white-ev slider , and if it ends up to flat add some local contrast.

Or… Use the tone equalizer with the mask set low enough they you can control that patch , and lower the tones there.

Or… Lower overal exposure and then use tone equalizer to bring it all up except the extreme highlights.


I did not look at the raw, but your darktable result has more contrast than the capture one result. Contrast pushes your highlights to be brighter, maybe you can try to reduce contrast in filmic to get a result closer to what you got with capture one.


Thanks @kofa! your second edit is what I was trying to achieve. :slight_smile:

@jorismak Deactivating highlight reconstruction and using the tone equalizer I was also able to get the highlights back. Thanks!


I just tested the new highlight reconstruction methods and the results are amazing out of the box without having read any manual :smiley:

_DSC7880_02.ARW.xmp (14.3 KB)

Who has to be thanked for that?


@hannoschwalm has done most of the work, with contributions from @garagecoder and me.


Great work guys

And we mustn’t forget @Pascal_Obry doing all those PR review/merge, without that there would be nothing. I’m constantly amazed how much he gets through.


A lot of thanks to you all!! I get very good results in all images I tested.