Basic semi-automated style for darktable 3.0/filmic

When I choose scene-referred filmic and exposure are already applied.

Do I use exposure again (new instance?) if I’m not happy with the result?

A bit similar with the *.dtstyle from post #1 - works very well. But do I apply i.e. local contrast or sharpening again?

Yes. History is the order you did operations, while the Active Modules shows the “pixelpipe”, or processing order.

Yes.

I adjust the same instance unti I’m happy. Or until I hit the 3EV limit, then it is time for another instance. That doesn’t happen often though.

2 Likes

The ‘3 EV limit’ is only for the slider, right click on it and you can type in other values (4, 5 or even 6 EV, if you want). It also allows for precise input values. And once you enter a larger value, the slider accepts that as a new limit.

2 Likes

Yes: history shows not the processing order, but the order in which the user added / adjusted the modules. For the initial import, the “user added / adjusted modules” does not really mean anything: there, the modules provide a starting point.

Now, for the adjustments:
As Aurélien has mentioned in one of the videos, the colour matrices are optimised for midtones, therefore it makes sense to set the exposure correctly for the main subject first. Filmic, in all its iterations, is a tone-mapping tool that then allows you to define a white and black point, and how the values in-between will be mapped to the display / output. It won’t move the traditional 18% (18.45%, if I remember correctly; Lab L=50%) midtone, but will shift the other values.


Download the gradient image attached to this post. Add a ‘live sample’ in LAB mode to one of the darker patches, then increase exposure until it reads 50. This will blow the highlights. Enable filmic, increase the ‘white relative exposure’ until the blown highlights are recovered. Play with the settings on the ‘scene’ and ‘look’ tabs.

5 Likes

I had a discussion about this on IRC a while ago and I think this is a misconception. Exposure adjustment is simple linear scaling of your rgb data, and a matrix profile is exactly that, a matrix. Matrix-scalar multiplication is commutative, so the order of exposure and input profile should not matter.
If you are using a LUT camera profile this is of course very different, and those may benefit from ‘correct’ exposure.

The ‘midtones’ you mention are probably not midtones in the sense of exposure, but saturation. Colour profiles are optimised around soft skin tones, because humans are very sensitive to that and if the skin tones are not right, the picture will look ‘off’. As a result, weird things happen with very saturated tones, especially blue ones as they are on the opposite site of the colour wheel compared to the pink/orange-ish hue of human skin.

Sure, the linear multiplication in exposure and the matrix multiplication are interchangeable; from what I remember Aurélien said that the matrix was just an approximation of the sensor response, and that approximation was optimised to produce the lowest errors around midtones. Therefore, it’d make sense to get exposure right, then apply the colourspace transformation, rather than converting from camera to working space, and then messing with the custom midtone point in filmic.

Not all input profiles are matrices, some people may use LUTs too (as a sort of ICC profile, or in the LUT modules), and for these profiles, anchoring grey first thing is important. Of course, for simple matrices, we don’t really care.

I ave made an icc profile using dcamprof which uses also LUT as you have mentioned so exactly what am I suppose to do? what do you mean by anchoring grey first?

Anchoring grey = set global exposure (exposure setting in exposure module) such that the image is legible and midtones are bright enough, with no care for highlights clipping (because they will be fixed later in filmic).

2 Likes

A recent suggestion was to drag the LUT module after filmic in the pipe to avoid conflict…so basically set the image tonally and exposure wise and then use the LUT on a nicely balance image…would you agree??

It really depends what are the assumptions that the LUT in making about the inputs (eg. input colour space), and if those conditions are met, what does the maker of the LUT claim it is meant to do (eg. maybe it can even replace filmic altogether, depending what is the purpose of the LUT)?

You must be aware of two different usecases for LUTs:

  1. correction of the colors captured by the camera to achieve „correct“ colors ( similar to the use of LUTs to process log footage in film)
  2. apply a look on a proper developed image

For the first usecase the LUT must be applied before filmic, for the second use case after filmic and some further modules …

1 Like

For sure and well stated…much like an ICC Lut vs a look or style…

I think a better solution is to enable the middle grey slider again in the options then you don’t have to go back to exposure…just drop it until the image is where you like …that’s my solution anyway…funny for me it often ends up around 9.45 which was a default in an earlier version of filmic…

I think its fixed and 18.45 now in 3.21 and that is why all the messing around with exposure and no grey slider…so you can enable the grey slider and use that as an option to going back and adding more exposure…

if you use LUTs in its default module order then it’s better to do the exposure compensation using the exposure module. Otherwise the LUT might be working on wrong preconditions.

I have a dynamic keyboard shortcut set up to adjust exposure. Then I can tweak it while I’m in filmic - I just hold down E and scroll my mouse.

Also good