Essentials of processing; tone curves vs filmic, and b/w conversion

I wanted to get a feel for how you guys process your raws, and when I say process I mean the bare minimum you may want to do on a photo apart from demosaicing, sharpening, noise reduction, lens corrections, defringing → that leaves mainly tone controls.

The reason why I ask is that I have been trying to use the new filmic module and while I see its benefits, at times it feels a bit cumbersome to use, and I found myself going back too often to the excellent article @anon41087856 wrote in his blog and start debating with myself uhm, what’s the dynamic range of my camera?.. how many stops should I have from black to whites, etc… Perhaps with more practice this will become second nature, but at the moment I’m stuck in deciding if I should persist in using filmic and use only that for all my photos, or should I use the simple tone curves.

And on the subject of tone curves, I’m also tring to understand if I “need” to use them in full RGB mode or in LAB mode working on each channel individually; once again I have seen contrasting views on this, people arguing against working in LAB mode and articles on the darktable blog showing on the other hand the benefits of working in LAB.

My internal debate comes also from recognizing that in the photos that I’ve processed until now, filmic does add a certain value, getting rid of sometimes nasty colors in the highlights for example, but is the final result so much better than using tone curves? Should I restrict using filmic only in photos with large variations in lightness, and perhaps skip it when the image is ‘flat’?

Perhaps it’s worth mentioning that my aim in photograpy is not to work on single images but build coeherent collections, sequences that hopefull will end up in book form sooner or later. I’m not a landscape photographer looking for ultimate sharpness etc, so my processing needs to be simple, minimalistic, and aimed at enhancing tones of my images, deepening shadows when I feel like it and sometimes lightly alter the colors (which is something that I’m able to do in Lab mode for example).

note: I’m aware of this thread but in here I want to specifically discuss about filmic vs tone curves and get a “final” answer, if that’s even possible.

note 2: I did have in the subject also “b/w” conversion because I wanted to get a feel for the most common/useful/used tool in darktable to do black and white conversion but maybe I’ll focus on the above discussion first.

1 Like

Ciao @aadm,

I am just now playing with the same things as you do :slight_smile:
and I sort of have the same questions as you, as well.
Which darktable do you use? I play with the 2.7.0-git,
but aurelien’s excellent article deals with 2.6, I believe.

My goal today is to learn more about conversion to b/w,
using darktable, and at the same time give more oompf to
the image. Here is the corresponding dtstyle: cglBW.dtstyle (7.8 KB).

Does it work with your version of dt?

Cordiali Saluti,
Claes a Lund, Svezia

Since version 2.6 came to be, I’ve replaced the base curve module with the filmic module for my basic processing style. I took the studio shot from the DPR test of my camera, opened it in Lr and in dt, and then tried to match the tonal values from the Kodak Gray Scale by using the filmic module in dt. Here’s a screenshot of my settings when I use the default (linear) input profile:

I also have a custom ICC input/camera profile which I made with DCamProf, which has a typical Adobe input curve embedded in it. This non-linear input profile requires a different filmic curve:

I find that these two styles are a great starting point for any well-exposed photo I take with that camera.

1 Like

Oh, I forgot to add that when using the default linear input profile, I had to use the exposure module with zeroed out settings but with the screen blending – this needs to be done to center the histogram when using linear profiles. One could alternatively use the unbreak input profile for that, I suppose.

Ciao Claes (parli italiano?),

re: darktable version – I’m cutting edge, I’m using the latest-latest 2.7.whatever directly from git and yes, Aurelien’s article deals with 2.6 but apart from one ‘saturation’ slider (can’t recall exactly what that is, I’m not on my home computer now) it is still a valid guide.

I’m sort of happy that I’m not the only one wrestling with these doubts! Let’s hear other folks and maybe I’ll add some examples to the discussion.

Thanks also for your dtstyle, I’ll check it out later and see how you work wiht b/w.

Nice to read about your approach. So you are 100% on filmic for all of your processing. How do you deal with some slight tweaks to colors in your images? If you do any, I mean.

I suppose your aim is to build something very “objective”, commendable. I wanted to ask what you mean when you say “linear input profile”.

Regardless of what I do later, ie filmic vs tone curves, i always start from deactivating base curve (well now I have a default so that all new images don’t have the basecurve applied) and try to extend the exposure to just avoid clipping pure black and whites, in line with Aurelien’s recommendation (the same thing was recommended also by Ming Thein in his processing tutorials anyway so I’ve been doing this as a standard procedure for some time now). Is this, i.e. flat basecurve, the same as “linear input profile” in your post?

The default input/camera profiles which ship with darktable are all linear in the sense that there is no embedded tone curve in them, but then you usually want to compensate for the perceived lack of brightness/underexposure by using something like the exposure module. The vast majority of commercial raw converters embed some kind of tone curve into their input/camera profiles (e.g. Lr, C1 or DxO PL) so that the user doesn’t have to center the histogram.

I don’t do much and try to keep things natural – I might use vibrance or toned-down velvia modules, or the a/b channels in Lab curve to add saturation. I might use color zones or color lookup table to modify some specific hues. I need to get down to the improved color balance module, but for now I’ve been only using it to fight with magenta vignetting that my camera sometimes suffers from.

Glad to see I’m not the only one experimenting with filmic :slight_smile:
I see filming advantages with difficult light conditions: high dynamic range, backlight, mixed sunlight / shade etc.
However I feel that for “normal” situations (indoor, controlled light like flash / studio) filmic has no significant advantages over the normal base curve / tone curve workflow, while being slower to tune and more difficult to copy/paste between images in the same shoot.
Also, sometimes base curve with exposure fusion gives me excellent results in HDR like images too.
Since I have Capture One and I like its colors, one thing which I did is to make a DT style which emulates tone and color of C1 for my Sony cameras.
The style uses the C1 color input profiles + unbreak input profile + a customized tone curve in order to exactly match the Kodak greyscale of the DPR studio raw file for my cameras, as processed by C1 into a tiff file.
I gets color and tones which are practically indistinguisheable form C1 standard output. In most of the cases, that is an excellent starting point for my workflow. Also, I believe that splitting the single base curve into a mild tone curve inside the C1 input color profile + a RGB tone curve applied later in the pipeline (just like C1 does) gives a better control and avoids hue shifts.

My two cents.
Marco

I do a batch process to make small proofs. Doing that requires a tone curve that works okay with a variety of exposure strategies, sounds impressive but it’s really short for ‘messing around’. Sooo… my batch tone curve is a linear black/white point ‘contrast stretch’, just to spread the data across the display range from wherever it sits. For proofing, it works well, and a lot of times is the only processing I do to a particular image.

I’ll review the proofs, and re-open any I want to work on. That mostly just involves a custom tone curve to meet my whims. Sometimes it’s just a lift, sometimes its an S-curve.

For my hack software I wrote a ‘tone’ tool that has a mix of curves: power gamma, log, gamma-log (see the thread on HPS), and a Reinhart tone map. I’ve messed with them in conjunction with all the discourse on filmic and linear, but I’m not ready yet to wicker any of them into the proof workflow.

My only experimentation with profiles is in camera profiles, investigating how to corral extreme colors, and lately “baking” white balance into a camera profile, again based on discussion here. The white balance thing looks promising, it doesn’t desaturate colors like white balance multipliers, but I really need to do some double-blind testing to characterize that.

Just back from a week of heavy shooting and experimenting using filmic 100%. I have a Fujifilm camera and despite earlier problems I am now able to use filmic without the black slider going ‘mad’ when I initially use ‘auto-tuning’. I then make further adjustments in filmic which provides good basic results with 99% of my images including the colorization.
I still need to use shad/high tool adjustments. I have tried very hard with filmic to boost the toe and shoulder but it lacks the flexibility and punch that I appear to need. I use local contrast gently.
Those 3 modules in the favorites tab have done most of the heavy lifting for this test series.
I look upon my images as composite pieces, so I then prepare to make local masked adjustments as are needed and then sharpen and de-noise.
There still are images that appear to totally baffle filmic, at which point I simply revert to a basic base-curve, tone-curve process which provides massive flexibility.
IMO filmic is a very useful tool that can provide good results under normal lighting conditions but I am thankful that I have other dt options as needed.

1 Like

If you look at @anon41087856’s workflow, it consists of several modules, not just filmic. E.g., see panels in A tone equalizer in darktable ? - #3 by Tobias, though the active tab isn’t showing…

Finally had a chance to try your cglBW style, Claes.

A bit too strong for my taste! On my Nikon files (D600) it doesn’t really work, the image becomes almost posterized. On Fuji raws it works as intended, I think, but it is too heavy-handed. Again, for my taste. I think it’s got too many things (fimic, levels, exposure, local contrast, lookup tables, etc) in it to be a truly portable style. Maybe show us some example of the intended effect? Maybe the way you shoot and expose is very flat so this style does not bring the image to the extremes like in my case?

Also: is filmic a portable module? I mean since it’s so fiddly (not in a negative sense obviously), maybe we should not make styles with filmic in it?

I would think that you’d want to redo your filmic settings for every image, unless you’re in really consistent light.

1 Like

Alright Mica, so you’re confirming my thought.