I have for some time used the new Filmic RGB module for all of my processing and slowly but surely my efforts appear too be getting better results. What concerns me is that the interface to this module does not have the same sort of user predictability as does, for instance, working with the tone curve.
If we ignore saturation for a moment, there are 4 basic adjustments that are made during this process.
First, the contrast slope is adjusted, following that the toe and shoulder positions are positioned and then the mid-tone point is placed. I am not suggesting that there is a special order in such a process, I simply think that making these adjustments directly would, in my opinion be more intuitive to most users.
I get the feeling that the current interface is driven by the background calculations rather than having the calculations being driven in response to the users input.
Don’t get me wrong … I think that this module is a great departure from previous strategies despite being a bit of a ‘black box’. For instance I often find that I get better results from a preset than I could achieve with my direct adjustments.
Have you watched @anon41087856 videos? There is like… 5 hours of explanation, so calling it a 'black box" seems a little unfair.
I’ve posted this elsewhere, but, here is the simplest workflow for filmic:
- Use the exposure module to push your histogram to the middle.
- Choose a preset that matches your scenes dynamic range.
- Tweak the mid tones, tweak the highlights, tweak the shadows.
I’m working with filmic since rc1. Saw Aurelien’s videos and adopted a workflow he described there (same like @paperdigits).
It’s written in detail in the darktable 3.0 manual. An early PDF is here:
But to be honest, I’m not experienced enough to give advice regarding filmic besides this workflow. I still find myself trying ‘some sliders’ with descriptions that sound like they could help.
And my highlights loose contrast in so many situations…
I also seem to crush my shadows and highlights more with filmic RGB than with the old filmic, for some reason. The old filmic with “intent: optimized” kept nice round shoulders for the s-curve, which made it easy to have nice contrasty mid-tones and still retain a good roll-off for extreme luminances. With filmic RGB, I find that I need to set the latitude way down to get that, and even that is not always enough.
I find it kind of a pity that the old filmic is deprecated. I liked it enough that I donated to Aurélien to thank him, and now it’s gone. (Or more accurately, it’s still there for backward compatibility, but it’s hidden from me.)
I’m not quite sure if it’s filmic module or not. When I shoot a landscape with a bright sky I bring up the midtones with the exposure module to prepare for filmic. Usually the midtones (below the horizon) get a good exposure and the sky is too high. After activating filmic rgb with that workflow described above my highlights often get flat.
Maybe I have to use a second exposure module for filmic preparation that brings down highlights if midtones and highlights are too far away from each other.
I guess we need a few good play raws.
Filmic RGB behaves better than the previous because the tone curve has been designed specifically. If anything is crushed, you need to enlarge the input dynamic range, thus rescale the black and white of the scene (just use the dynamic range scaling slider, in the bottom).
You need to keep in mind that filmic is a technical module, not an artistic one. Your camera has 1.5× to 2× the dynamic range of your display. Filmic does a 3 points mapping (white, grey, black) of the camera dynamic range (scene tab) to the display dynamic range (display tab). The critical settings are the scene ones. The look settings are designed to be portable and preset-able for styles and batch editing, provided your scene parameters have been adjusted correctly depending on the actual input picture.
Then the tone curve is just there to define a high priority zone for the remapping (that is the latitude), that can be enlarged (if contrast > 1) for better local contrast in midtones, and allow some slack to the extreme luminance, and blend everything smoothly.
For artistic contrast corrections, you get the colour balance and the tone equalizer modules, even the RGB tone curve.
Trying to use filmic as a tone curve is not good practice since it will put too many constraints on it. Think of it as the companion of output colour profile : one does the colour mapping to display, the other does the dynamic range mapping. Both ensure proper adaptation of the image for whatever display.
I think to make the best out of filmic, you really need to split your edits (divide and conquer):
- Set up the luminance thresholds of your scene, e.g. the bounds of the dynamic range (white, grey, black). White and black depend only on the technical parameters of the scene (max and min RGB values in the picture), and grey depends on how much you want to decompress shadows (reminder: if you shot a colour checker on scene, you can directly use the colour picker on the grey patch).
- Set up the the look, e.g. how much you are ready to compress extreme luminances to keep good looking midtones. Note that Filmic RGB needs a lot less contrast than the previous one. Keep the latitude as wide as possible whide avoiding clipping, and slide it up/down depending where your important details lie in the luminance range.
- Don’t touch the display settings if you have a regular 8 bits screen SDR with gamma 2.2 following ICC specs.
You will find that for a specific camera, your look settings will consistently be the same from one picture to another, according to your taste. Therefore, put them in a preset, load them by default for new pictures, and all you need to care then are the scene parameters.
But you also need to wrap your head around what steps in your editings are driven by technical constraints (display gamut, display contrast, human vision, etc.) and what steps are driven by artistic choices (colour, brightness, etc.). If you get aware of that and carefully divide the technical job from the artistic one, you will fight a lot less with the software.
Ah, I see, thanks. Is there any practical difference between using the dynamic range scaling slider and modifying the white and black points directly or are they two ways of achieving the same end result?
No, it’s exactly the same thing, the dynamic range scaling is only a shortcut that ensures you keep the balance white/grey and black/grey identical while you rescale.
Great, thanks again for your help and clarifications.
Thank you Aurelien … this is another way for me to look at the module’s work flow. When I look at your presets I notice that the ‘look’ sliders are all in the same position regardless of the DR; I am assuming that this works well when there is a well balanced shadows with highlights with ‘zone 5’ nicely centered.
All of your presets maintain that central position for zone 5 whereas if I use the ‘eye droppers’ or change the relative-exposure sliders that position is re-calculated and moved.
Hopefully I do have a better understanding now of the operation and my results will be less hit-and-miss.
Thank you for the very fine work.
I have posted it in the files of the DT FB forum some time ago…if anyone wants it from there…
The presets use the same look settings but not the default…I recall one has 22.4 % dynamic compression and I think the default settings used for the auto have none…I think there are other changes…I noted this when I got different results from Auto when I started from the default setting rather than the closest preset to my Dynamic range…
I think this is the slider I find myself playing with the most to get the right look to the scene, and the one which I have most trouble ‘getting right’. After adjusting the grey I often end up playing with the black/white points as well. I’m not convinced this is the ‘right’ way, but it gets the results I want.