Filmic usability -- my usability "study"

Soo, I said I was going to post a video of me trying to use filmic to point out how I think it could be improved (and why using it is costing me a lot of time at the moment), somewhere in this thread

Turns out it’s taking me some time, and my spare time has just become a bit rarer, so here’s the first part.
This is not the finished image, just the halfway edit after setting up exposure and filmic to be probably kinda not terrible, before starting with local adaptations:

_MG_0653.CR2 (20.6 MB) _MG_0653.CR2.xmp (14.1 KB)

And here’s a video of me getting to this point:
filmic_usability_1.mkv (58.5 MB)
It got a bit longer because I also took the time to explain a bit more what I’m doing and why, despite some editing. Turns out text is still easier to edit than rambling explanations. And I need to learn to ramble less. It’s an ongoing effort…

I’ll try and summarize the more important points:
1: overshooting roll-off curves are bad because they break the link between moving a slider and the visual effect on the photo – fortunately a solution for that is already in the works.
2: But because the filmic curve stays where it is when changing exposure, filmic always needs to be re-adjusted after changing exposure, at least in cases where the user cares about (not) clipping highlights. Which in turn means that the direct visual feedback when moving the exposure slider is limited. That means reconsidering the exposure setting at any point after the initial setup of filmic is time-intense.
3: It’d be nice to be able to manipulate the upper and lower part of the curve completely independent of each other (except of course that they are linked by the middle-grey contrast)

There are some more technical details in the video, but I’ll discuss them in the other thread, in an attempt to separate the discussion about how to find the best settings in filmic efficiently, and how to adapt the UI to make it more straightforward to get to that point.

I’ll add the remainder of the photo processing here, together with the final XMP, once it’s done. I hope it’ll take less time than the first part.

I know there are some people who approach it completely differently, and I’m curious how they would do this. Do you simply not try to use filmic as much and directly proceed to local adjustments? Do you always have a well-defined “subject” in your scene, adjust exposure to put it at middle grey, and then work around that? Or do you change brightness of the mid-tones in display-referred space, after filmic has been applied?
Or am I missing some important strategy?


Why don’t you post this as a play raw and ask people to edit it and not compress the history stack before posting the xmp. Then you could see step by step how they approach it.

I always set the exposure before I mess with filmic. The exposure module comes before filmic in the pipeline, so any change to exposure will effect filmic.

Another thing about filmic. If you image has low dynamic range, you are sometimes better off shrinking the dynamic range instead of opening it up. You can always ask filmic what it thinks of the image by using the color pickers to set the black and white points. If you don’t like it, then reset it and adjust to your liking.

In the image above I would set middle gray from the hillside between the path and the creek. Then I would adjust filmic black point, white point, contrast, and saturation to get something to start with. After that I start editing. If you keep your edits below filmic in the pipeline, then filmic tries to keep everything from being under or over exposed. Sometimes I move modules above filmic so I get the “unfiltered” result. That’s really a matter of experimentation. If the module doesn’t give you the result you want when it’s below filmic, try it above.

Setting exposure and filmic normally takes me less than a minute and never more than 2.

One last thought. Don’t try and develop your image to fit a histogram or a curve or the exposure indicators. Develop your image to be what you want it to be. If you want to crush the blacks because that’s the look you want, then don’t worry when the under exposure indicator lights up.


You may have addressed this in the video but the thing that stands out to me instantly is there are too many changes to the exposure module.

My usual process is : set exposure for the midtones, ignoring the highlights and shadows as per the manual, set filmic, tweak exposure but only if it needs it.

It’s rare that I then need to go back the exposure module again.

p.s. I like the idea of providing a video.

1 Like

That is a great idea as I think it would help others including myself.

1 Like

The first thing in my approach is: What do I want to achieve ? What is my goal ? Where is my “subject” ? For me this picture contains roughly 2 main subjects: The sky and the mountains.
For me the mountains are more interesting, as the sky is heavily blown out and lost most of its detail. So, the mountains would be my subject and thus determine my “middle grey”. I would lift up exposure until this subject is clearly visible. (I never measure middle grey as you do in the video.) Next I would add some contrast and maybe saturation. And that’s it.

_MG_0653.CR2.xmp (13.0 KB)

I watched your video and it is not clear for me where your goal is. Do you want to make the landscape visible, or is your interest in showing the structure of the clouds ? Are there any other parts of the image you want to make visible or lead the viewers eye to ?


Getting to this point is usually a matter of a couple of minutes’ work for me. I use the ISO12646 color assessment mode (bulb icon at the bottom of the screen or Ctrl+B) to set mid-grey, then filmic (still in color assessment mode), working through the tabs from left to right. Then I add some local contrast and play around with saturation until I’m happy.

Here’s my two minute effort in darktable 3.4.1:

_MG_0653.CR2.xmp (13.3 KB)

Finally some local edits to darken the clouds (tone equalizer) and lighten the dark patch in the bottom right (masked exposure):

dt 3.5.0 treats the clouds a little more forgiving :wink::

_MG_0653_01.CR2.xmp (11.5 KB)

And I forgot to say: I only tuned the black and white relative exposure in filmic, everything else in this module is left untouched .


Perhaps I am just lucky but I found that after increasing exposure by 0.5 to 1 stop, the default filmic delivers a pretty good starting point, remarkably close to OOC jpeg for a couple of different cameras. (better with preserve chrominence set to no IME). I tend to leave it alone unless it is an extreme photo and use tone equaliser to adjust shadows, highlights etc.
My only difficulty is that sometimes skies are too desaturated. If you increase the latitude to increase the saturation of the highlights then the curve starts to reverse. Perhaps this is sorted in later versions.


Perhaps that means you have set the white point too low, in effect over-exposing the blue sky. As highlights cannot be coloured (*), there’s a progressive desaturation towards the highlight end of the histogram.

*: See the articles by Aurélien Pierre for the theoretical background

1 Like

First of all, thank you for taking the trouble to clarify your comments regarding the use of filmic with an example in a video. So one can understand your approach and emerging concerns well.

It depends on the photo. In this article I have talked about it and shown a few examples:

I think that you always have a subject in the photo, otherwise you will not make that photo. But this does not mean that you have to have this subject in the “middle gray”. This is a technical and very rigid term and serves only as a rough orientation. What is “well lit” is subjective and depends on many factors.

In this example, the warning in the foreground on the ground between the rails and the part of the street in background, illuminated by the evening sun were important for me because they indicate a strong contrast between the danger in the time of the pandemic and the nice inviting weather for sightseeing:

Here it is not a middle gray that is important, but the relationship between these two elements, which must be placed in such a way that both are easily recognizable, but at the same time, without the “reality” of the scene suffering much.

What is necessary for brightness value, you can recognize very well with your own eyes, without relying too much on any correct “measured values” in the processing.

This is what an unprocessed raw file looks like:

If the dynamic range requires it, of course! The main thing here is to adjust the brightness locally, which should make the main subject stand out better.

If your subject is well lit, like in the studio, you can finish the processing with filmic. But this is rarely the case when you shoot outside. You don’t always have ideal lighting situations.

Accordingly - as far as brightness is concerned - it doesn’t make much sense to always see the entire photo, but rather to see the elements in the image that play the role.

The process with filmic is very similar to the process when you previously switched off the base curve and have more control over the processing. The advantage of filmic is that in addition to the possibility of being able to deal with dynamic range very flexibly, by imitating the analog film, you already have a solid basis for further processing of the photo.

And depending on this dynamic range, you can immediately achieve good results with filmic, or you can use it as a tone mapper and continue your work with other modules.


Here is my take on your pic and the xmp with all the steps

_MG_0653.CR2.xmp (22.2 KB)
I spent most of the time in recovering highlights (which are blown, unfortunately) doging and burning, enhancing contrast, masking, color grading etc, but I touched exposure only once.

1 Like

What I’m about to describe isn’t darktable per se, apologies to all who are focusing on its knobs and dials. But, after my own wrangling with parametric curves such as filmic, I’ve come to the conclusion that sometimes, you just have to go rouge…

I downloaded the raw and tried both my four-parameter filmic and the two-parameter doublelogistic, but I couldn’t pay homage to both the dark and light regions with either. So, I reverted to a two-step process: 1) a loggamma curve to spread the data out, followed by 2) a control-point curve to shape the tone transitions to my whims. Here’s a screenshot showing the image at the control-curve tool:

Tell a parametric curve to do THAT.

For completeness, here’s a screenshot of the tone curve pane showing the loggamma curve applied to the linear data:

I use the HEVC version of the ARIB STD-B67 algorithm, described at


Red? :slightly_smiling_face:

1 Like

Rogue! When spellchecker cannot help you… :smile:


My take. I tried to lead the viewer through the image using the path and the water, plus lightening the valley to draw attention.

_MG_0653.CR2.xmp (29.2 KB)


Yes, that’s the way you could see it. The play of light on the left slope of the mountains combined with the swiveling path of the stream that draws its course from the shining mountain behind is what makes the scene special. The dramatic clouds enhance that even more:

Here is a softer…

_MG_0653_04.CR2.xmp (25,9 KB)

…and a much more dramatic version:

_MG_0653_05.CR2.xmp (31,3 KB)

Too bad the clouds are overexposed. For this you lose most of the time in editing.


@Mister_Teatime I have another suggestion. Watch @s7habo’s Editing Moments video series. Don’t watch to just see how he transforms an image, but watch to see how he uses the tools and what he uses them for. There are a lot of videos that are about this is what the tool does and this is how it works. Editing Moments is about how to use those tools to effect a part of an image and how to stack those tools and edits together to achieve a finished edit.

I really like your clouds, so I’m going to peek at the xmp. I’m betting it has something to do with channel mixer (color calibration) and an overlay blending mode :smiley:, plus a few other tweaks


Filmic really is quite amazing, the new undocumented “add pug” feature works really great.

_MG_0653.CR2.xmp (74.6 KB)


Yes, adding the fun back to Play Raw! Reminds me of [Play Raw] Sacre Coeur → Backlight challenge.

1 Like

what? I’m pretty sure I did not compress the history stack…

Also: Feel free to edit it, but the point of this was only partially about the result but mostly about how to get there quicker, how to know where to go in the first place, and how that could be accelerated. Which is why I felt that it wasn’t quite a play RAW kind of post.