My question comes about from the effect of the skew slider. Moving the skew slider to the right can give a more dramatic sky by adding contrast to the clouds, but this can have a negative impact on the shadows. Now moving the slider to the left can increase the contrast of the shadows and improve the look there but to the detriment of the clouds in the sky. Is there any reason I should not use two instances of sigmoid and localized masks to bring out the best sky and the best shadows?
Not sure why⊠use the tone eq⊠Sigmoid and filmic are global tonemappers used to manage dynamic range⊠Of course you can but I donât see it as a tool for doing local edits
No, I donât think so. I may have done this myself on a particularly tricky image⊠canât remember now.
Only issue could be if you end up with overlapping masks - you do get a bit weird tonality with âdouble sigmoidâ.
The documentation to sigmoid sais:
" only use one display transform
Never use sigmoid together with another display transform module (i.e. filmic rgb or base curve)."
I would believe this includes sigmoid itself.
TE ought to be for this kind of tasks.
EDIT: The manual also says: âYou should try to accomplish the majority of your processing using modules in the scene-referred section of the pixelpipe and not rely on the display transform module (sigmoid, filmic rgb, base curve) to do all the work.â
But if the two instances are masked off from each other, than the images as a whole only has one instance, no?
Hm, I forgot Terry referred to âlocalized masksâ, but wonât anyhow those image data that comes out of the first instance be transformed to display referred, and as such have a wrong format for input to the second instance?
1 Like
Yes, I also thought about asking him.
I also thought that to avoid any risk of problems in a two instances process one should also probably need two perfectly complimentary masks, which should mean no soft edges â which likely will create some kind of unwanted lines in the image, wouldnât it?
1 Like
It seems to look ok, from a quick test⊠this is two identical instances of sigmoid, split vertically roughly down the centre by a mask.
Thereâs a small funny transition at the border near the top, but apart from that seems ok.
This is a zero feathering mask - both instances using the same mask just polarity toggled.
All files licenced CC0.
IMG_0188.CR2 (17.3 MB)
IMG_0188.CR2.xmp (13.5 KB)
MMm. I tried feathering the single mask from above example, and sure enough, somethingâs not quite right.
I think @s7habo once used two instances of filmic with an edit. It was from a Fuji camera, if I recall correctly, shot in a forest, with a huge dynamic range. Iâm not sure if there were even masks. If I can find the image, Iâll link it here for inspiration, or maybe Boris knows what Iâm talking about, and will link it himself.
I donât think that was me. I have played with several instances of filmic but mostly in conjunction with blend modes and that would be an atypical edit for me. But, maybe youâre right. Who knows, Iâm getting old and forgetful. 
Now I will say, knowing the restrictions of the sigmoid, you can tweak things in post-formation (after tonnemaping).
Here is original photo with smooth preset of Sigmoid:
You can, for example, first adjust the contrast and skew in the sigmoid until you have nice contrasts in the shadows and mid-grays:
Then you can use ton equalizer to reduce the brightness of the area in highlights that is later in pixelpipe strongly influenced by sigmoid:
To adjust contrasts in highlights only, you can bring color balance RGB module over the sigmoid and use briliance sliders:
You can use a new instance of tone equalizer without detail preservation for the same purpose (simple tone curve preset):
You just have to bear in mind that you no longer have gamut protection after sigmoid and have to be correspondingly careful. For example, I use the waveform as a guide to how far I can go.
5 Likes
It was. I remembered it because it was so unusual:
2 Likes
OMG, yes!
Thatâs how it looks now 
:
I misused the second instance to increase the contrast with multiply blendmode. Please donât do that!
2 Likes
Indeed. But then it means thereâs a problem with backwards compatibility.
1 Like
Short answer is no.
Sure you could have a binary mask and invert that but you would have get a weird seam between the two.
You could however do it if you are willing to blend the two results externally from darktable:
- Make two separate edits.
- Export and open Gimp or similar.
- Blend them there as you like.
You can apply multiple global tone mappers in parallel and merge the results but not sequentially. darktable only does sequential processing so you shouldnât use two tone mappers in darktable. Node based editors such as vkdt opens opens up for this kind of workflow!
3 Likes
I think this is not a good idea at all. Itâs much better to use sigmoid OR filmic to get to a base starting point and then use curves and other tools to get to where you want. If you use RGB curve instead youâll get much more precise control over your shot.
indeed.
consider this graph with two film curves, i.e. 01 exposing for the foreground:
and 02 exposing for the sky:
thereâs a special blend mode for exposure fusion, it is somewhat parameterised with the opacity parameter that instructs it to keep more of the low or high frequencies.
resulting in this final render, ymmv:
itâs often a good idea to set blend mode to over and set opacity to 0 or 1 to see one or the other filmcurv/whatever module have you in between.
4 Likes
I am glad I asked this question. I had never before used two instances of sigmoid, but I had an image which did benefit from doing this. But it seemed wrong to do this as surely the input going into the second instance was already transformed by the first instance. But it looked good. Hence my question.
I had fallen into a trap and depended too much on the skew slider in sigmoid to achieve the desired look. I was neglecting the contribution of the tone equalizer module to achieve the shadow/highlight balance that I desired. The examples shown by Boris showed me that I could achieve the effect I was desired by using the tone equalizer module as it was intended. Boris also demonstrated how the color balance rgb module could further improve the contrast of the highlights.
Thanks all for your great responses.
1 Like
Sigmoid has that nice fall off to white but it also in the process really destroys the detail so I think the tone eq is the way to go to offset that. Skew can help a bit but actually dropping contrast seems to be the best way for the some of the highlight details to come back a bit and then the look changes again and may be less desirable to some. Filmic does it as well but to my eye much less⊠I usually have tone eq local contrast and diffuse and sharpen if the mix and often I will not even use one of the tone mappers unless I have to because of DNR⊠tone eq and rgb color balance with local contrast and diffuse and sharpenâŠmaybe dehaze will provide a result with more details when you can go this route because you are not needing to manage DNR with a tone mapperâŠ
2 Likes