Thank you very much for the writeup. By now I must have spent the equivalent of multiple days learning darktable, and your text still gave me many useful suggestions and ideas.
I think it would be a great addition to the darktable manual, as an advanced tutorial.
Good idea. I just need to clean up the text a bit.
Thank you so much for this. As a darktable user I felt I was getting left behind with the recent changes and confused with the lengthy posts about tweaking this and that or why I shouldn’t use my old favourite modules.
I have been using ART and Filmulater of late but I will try out your speed technique.
Your welcome! But I do very much recommend taking your time and learning the theory behind the new scene-referred workflow. Check out Bruce Williams videos if you haven’t seem them already: https://www.youtube.com/user/audio2u
Also, Aurélien Pierre’s videos are also good. https://www.youtube.com/channel/UCmsSn3fujI81EKEr4NLxrcg
I am not ‘style experienced’ and never did a re-order pixelpipe. So I tried this with Fuji 400H4++. Does that look correct?
Is there a specific reason for using sharpening? I usually do contrast equalizer and lift the chroma curve a little. This is based on how I understood your entry post. I do get quite a strong purple tint though.
40s-restack-fuji400H4++.dtstyle (2.7 KB)
Which one is it? I don’t see that exact name,
My reply as I understood your questions:
There is no 100% right or wrong here.Could it be that there is a typo in the above values? Because setting them in filmic with darktable 3.4.1 leads to an overshooting tone curve.
Nope, those were the default values for my style at the time.
I wrote: “My custom style gives me a reasonable starting point, but seldom an immediate perfect result.”
The creator of the Filmic module stated explicitly that there can be no default value that works all the time. And I agree. You have to made a set of styles that work for you, and then adjust the values for each picture you take.
What is a decent starting point depends on the type of lighting conditions you have, what kind of artistic transformations you apply, the dynamic range of your camera and the lens you’re using.
Btw, I used Darktable 3.2.1. But I don’t think that matters in this case.
How does these styles transfer between different cameras? Do you need to create one style per camera (brand?) or do they generalize to all? Would make a lot of sense to ship this + a couple more styles as default darktable styles to make different editing techniques more discoverable!
(my current understanding is that it won’t transfer well between cameras which is a pitty if true)
How does these styles transfer between different cameras? Do you need to create one style per camera (brand?) or do they generalize to all?
I only have only tried my styles on 3 different cameras (one Sony and two Nikon). Usually, the quality of the light has a larger effect on how much adjustments I have to make, than the brand of the camera. For example, I took some photos of fireworks, and for that I make a special fireworks-and-night-sky-style.
Would make a lot of sense to ship this + a couple more styles as default darktable styles to make different editing techniques more discoverable!
Personally, I don’t think so. Darktable is for people who want to learn for themselves. But the Darktable handbook could use more hands-on guides as complement to the more technical documentation.
But this is open-source. You could always file a bug report, and see what others think.
I do understand that any default for filmic will not fit perfectly most of the time.
My comment was related to the purely technical observation that you were using an overshooting and therefore clipped tone curve. Your old curve is actually mostly equivalent to the following non-overshooting one:
This means an input dynamic range of 8.43 EV which seems a bit on the low side, even for a generic default. For example, filmic’s default parameters correspond to an input DR of 4.4 + 7.75 = 12.15 EV.
I’m thinking about these things because I wonder what good default DR settings would be. For example, for Olympus OM-D cameras, I am able to roughly reproduce OOC JPEGs by setting
But I can also keep filmic’s default, and then use the local contrast module to increase contrast. I have the impression that the latter solution tends to yield better results. (=Subtle local tone mapping.)
I find that stretching the histogram to the edges with filmic yields shadows that are pretty flat. Local contrast helps bring back that shadow contrast and is recommended as an addition module by the documentation for exactly that reason.
Now I see! What your asking is “why did you reduce the dynamic range so much”. A valid question, especially considering that my camera (a Nikon D750) has a maximum dynamic range of 14 EV.
The short answer is: “Because I like how it looks”. 
I’m in the same camp as Jamie Windsor on this one. https://www.youtube.com/watch?v=zCOI7bYDPrM
I’m fully aware that sometimes my highlights become clipped with this style. Most often it is the bright sky that loses details. Then I increase the dynamic range, use the tone equalizer or apply a graduate filter effect (very useful tool).
One of my absolute favorite thing with the Filmic RGB module, is that you can reduce the dynamic range and still retain smooth transfers between midtones and highlights. Unlike the old, bad times when we only had the base curve module.
My advice to you is to set the default to what you think looks good, and not thinkt too much about the numbers.
That was a great video - thanks for linking it. I agree there seems to be an overwhelming trend to always lift shadows and lower highlights which essentially evens out the light so everything can be seen in detail. Can be quite boring and often distract from the point of focus.
Thanks for the link to the video. I totally agree that there’s no point in always trying to display the full DR. And I also hate washed-out edits that reduce contrast in order to squeeze too much DR onto a SDR display.
However, I was left wondering whether your default does not go too far into the opposite direction. (I mean for a generic default, of course increased midtone contrast is a valid artistic tool.)
When one sets your old curve (white @ 2.63 EV, black @ -6.13 EV, contrast @ 1.555) in filmic rgb and regards the “Ansel Adams zone system” plot that filmic started to provide with darktable 3.4, one sees that the 2 EVs around scene middle gray are mapped to almost 3.5 EVs around middle gray on the display. In other words, with this curve contrast of midtones is exaggerated by almost a factor of two compared to reality.
Of course you are free to choose this as the default that you prefer, but the example pictures that you uploaded above do not look like they were using such high midtone contrast.
This is interesting, would you care to elaborate? I thought that any tone mapping that filmic allows can be in principle also realized by adjusting base curves. The advantage of filmic would be that it makes it easier to realize sane tone curves. (There’s also the advantage of the linear workflow, but that’s a different thing.)
Or as the author of filmic puts it:
It is important to understand that this module is roughly a combination of the unbreak color profile module in logarithmic mode and the tone curve module, with additional sweeteners, for a specialized use.
Thank you for the clarification. You wrote:
However, I was left wondering whether your default does not go too far into the opposite direction. (I mean for a generic default, of course increased midtone contrast is a valid artistic tool.)
A valid question! I have assigned dynamic shortcuts to all the modules I use most often. Therefore, I do not have to look at the parameters when adjusting my edit. (With dynamic shortcuts editing pictures becomes more like playing a video game than solving an engineering problem).
I usually make some minor adjustment on every picture, after applying my favourite styles. If I notice that I have to make a lot of adjustments all the time, I update my style. Otherwise I leave it at it is.
But your question got me curious on what filmic parameters I usually set. So I looked at a sample of (N=104) of the highest rated images I’ve taken the last 3 months.
As for the white relative exposure, 2.63 (the default of my most used style) was the most common value. But I finished a large number of images with 1-2 EV higher white relative exposure (WRE) than my starting point. But very seldom did my final WRE exceed Filmic’s default value.
So you’re right that 2.63 was a bit on the lower side. But not so low that it bothered me.
When we look at the black relative exposure we see that I almost never deviated from the setting in my favourite style.
And the final dynamic range was almost always lower than the Filmic default.
So what is the message I want you (plural) to take home?
This is interesting, would you care to elaborate? I thought that any tone mapping that filmic allows can be in principle also realized by adjusting base curves. The advantage of filmic would be that it makes it easier to realize sane tone curves.
Yes, the basecurve and filmic modules have the same purpose. But I found it very difficult to make any adjustment to the base curve so I always left it at some default value. I made an comparison between the modules in this thread:
