new to darktable and editing raws in general. something i came across a couple days ago is that you should not use more than one tone mapper, but search through here as much as i could i wasn’t able to find too much an explanation why you wouldn’t want to do this.
for context i tend to use both agx & filmic (in that order) in my own edits. it’s perhaps a creative choice? i use agx to get 80 to 90% of the look down and filmic for the rest. i had initially started out only using the filmic module, but i find agx’s tone shaping capabilities extremely powerful & flexible. it’s also so useful for nudging colors where i want them. on the other hand, i like the little icing-on-the-cake tweaks i can make with filmic—but especially for the highlight reconstruct, since my style is heavily film emulation oriented.
to use a perhaps flawed analogy, in music production i usually use two tape emulations in the master chain; each tape plugin sounds & behaves in a different way and in their combination i am able to achieve something i otherwise wouldn’t be able to with each individually.
what’s the risk or what’s undesirable about using a combination?
The main purpose of tone mappers is to squeeze the available dynamic range into a narrower band. They offer a bunch of controls to help you do this in an artistically satisfying way. When you chain two tone mappers, the one on the top (i.e., the one that comes later in the pixelpipe) does not have access to the wider gamut that is available in the scene-referred part of the process.
That said, If something works for you and you achieve a result that you like, then godspeed go ahead and keep doing it. There is no right and wrong when it comes to implementing your vision, what matters is that you are happy with what you get.
I would argue that probably you could achieve the same results using a more conventional workflow, but this is not necessarily the case (and, again, it does not really matter as long as you are happy with the result). Why not posting a playraw with one of your edits, so that folks can have a go at it?
The tonemappers adapt the dynamic range by compressing the shadows and highlights to maintain a given contrast in the midtones. They also encode intensities to a log scale ( i.e. I \rightarrow \log_2(I) ). You may not want to do either of those two steps twice…
But if you get the results you want, then the method is correct…
Tone mappers are curves, but, unlike traditional display-referred curves:
the input is not restricted to 0…1 (it’s unrestricted in sigmoid, user-defined in filmic and AgX)
they may come with some mechanism to manage colours in some way (either keeping the hue constant, protecting saturation, like filmic does, or introducing intentional colour shifts and desaturation, like sigmoid|smoothAgX do).
You can apply two of them, they can take 0…1 input as well, but then it’s more like an effect curve, not tone mapping
There is no “risk” (in the sense that it would make your computer explode), what you see is what you get, if you like the results when go for it.
It’s just that these tools were not designed for that purpose. But it’s OK to use tools for purposes outside their design, especially for creative effects. Note however that you may be able to reproduce similar outcomes with one tone mapper + tone equalizer, and potentially save them as a preset.
The risk could be that the result looks undesirable. If the effect of stacking tonemappers is to your liking: Go for it. I like your analogy to digital audio processing where “everything is allowed”.
thanks all for the replies! i think this supplies me with what i needed to know.
that makes perfect sense, and honestly that’s specifically something i would want. considering film’s reduced dynamic range when compared to current digital sensors, smushing it down into a flatter look is explicitly what i set out to do; however, as you say later, maybe there’s a better way to achieve this? there is a bloom module afterall, so i could try playing with that.
i’ll try not to be too self conscious and maybe give that a shot
glad the analogy worked for someone! my general approach to anything creative—music, writing, photos, etc.—is just to fuck around until i find a way to express what’s in my mind & heart. if not for experimentation how else would we find new means of expression?
If you want to squish the output range of an image you can use the target white and target black point controls in AgX under advanced curve parameters. It is simple, effective, and saves having to add an additional module that might be doing more processing beyond what you need.
Bloom is more for increasing the size of bright highlights in a pleasing and filmy way. For example, it will cause a headlight to enlarge and blur. This also results in a reduction in contrast/detail in those bloomed areas. But, blooming does not change the overall white/black point.
Reducing the dynamic range of the image can be done in two ways, reducing global contrast (color balance RGB, or the contrast sliders in AgX, Sigmoid, etc.), or by changing the target white and target black in something like AgX, which I mentioned in my other response.
The reason to use the target white and target black parameters is because you can maintain the same relative global contrast of the image while compressing the image in a very nice way.
For a film-like tone curve in AgX, I typically use the below settings:
Contrast : 3.25
Shoulder power: 1
Toe power: 1.55
Toe start: 20%
Target White: 97.5-95%
Shoulder start: 40%
Target black: 0.25-1%
i guess i didn’t provide enough info so far or maybe explained myself poorly, but i’m definitely aware of both of these points & do already implement this as described. again, i think agx’s tone shaping capabilities are incredibly powerful & flexible (at least in my so far limited experience). i use agx for all of the heavy lifting and getting the image mostly where i want it and then filmic to massage it the rest of the way there. the reconstruct section of the module has a lot of flexibility in terms of how i want to build that bloom, both in it’s range/size but also in its quality & color. i guess that’s why i use the two in conjunction instead of having reached for the bloom module specifically.
when i get a chance i would like to try out your suggested settings in agx to see what that does to an image and where that sits with how i’d normally edit something.
I have used base curve fusion and AgX which effectively I guess is two tone mappers. I also like AgX with some non-raw files which technically don’t need a tone mapper.
If it looks good it is good, but one tone mapper should suffice for most images and other tools can then help achieve the final look. A playraw of your technique would be interesting.
The effective dynamic range of an image can also be reduced by moving the highlights and shadows levels closer to each other by either of:
a) applying a linear compensation curve through the mid gray point in Tone Equalizer with preserve details = no, (typically by setting +2 at the shadow end and -2 at the highlight end),
b) increasing the shadow values of the perceptual brilliance grading, and reducing the highlights in Color Balance rgb.
Yes that is true. I place those in the category of global contrast adjustment, but I should have mentioned for completeness. I actually use those two methods often haha
The tone mappers don’t change the dynamic range of the image, they map (part of it) to the output range. Which part is determined by a black and white point in filmic and Agx (sigmoid theoritically never reaches black or white…). And they all do the mapping by lowering the contrast of shadows and highlights to maintain a given contrast around middle gray.
That may sound like nitpicking, but it has some consequences when you have a high dynamic range in the starting image. By reducing that dynamic range before the tone mapper (e.g. with masked exposure, or tone equalizer) you have much better control over which part of the tonal range gets compressed. That makes it easier to keep local contrast where you want it (and let it go when it’s not important).
Any change to the ratio between max bright/max dark is a change to the dynamic range. If I toggle AgX or any other tone mapper on and off, those will usually shift in my image, unless I have my image compressed to all be middle grey (the usual default midpoint that everything shifts around).
Since you can also directly change target white and black points, that also counts as changing the dynamic range.
While changing contrast does not change your white/black points, it can shift the “effective” white or black point. For example if the image data is very compressed coming into your tone mapper (not even touching the white/black points), and then the contrast is increased, the brightest and darkest areas of the image will expand and get closer to the white and black points set in the tone mapper.
Your advice on how to have better control over tone compression is totally correct, but tone mappers do change the dynamic range of the image, as far as I can tell.
Well, yes, but only in that they force the dynamic range to 0…1 on output.
But black point and white point do not change the dynamic range, they “just” change which part of it is mapped to intermediate values in the output (where the range is fixed to 0…1!).
And (IMO) you should not use the tone mapper contrast controls to change the apparent dynamic range. In the example you give about an image covering only part of the input range, changing the contrast can fill more of that range. But:
the mid-tone contrast will be increased (potentially by a lot);
the change in contrast for the highlights and shadows will change much less, and may even be diminished, depending on the exact mapping curve you end up with (that curve also depends on the contrast setting).
So I think it’s better to adjust (mid-tone) contrast before you reach the tone mapper, and if needed also adjust the dynamic range and local contrast, and then only do final adjustments in the tone mapper.
But it may be mostly a conceptual difference, in how you approach editing.
Yes that makes sense. I would generally say that the more you can do before the tone mapper the better, because you will have more control at each step of the way and you will be able to edit linearly through the pixelpipe.
