RAW developer: Darktable leaves me completely in the dark
RAW converter Darktable is open source and non-commercial. I’d like to like it, but I just don’t understand it.
RAW converters are often expensive. Some – such as Adobe Lightroom – are only available with a subscription. Others, such as Capture One, can be bought outright, but upgrades are costly – and new cameras often require a new version.
Non-commercial, open-source software could be the perfect solution. Not only would you save money, you’d also be spared annoyances such as mandatory registration, advertising e-mails or prompts for paid extras. As far as I know, Darktable is the most powerful RAW converter in the open-source space. It runs on all reasonably common operating systems.
Darktable’s difficult
I’ve wanted to write about Darktable for over a year. Here it is – although not quite as planned. Instead of a review, this is an account of my experience without a rating. The reason is that I still don’t fully understand Darktable.
Most RAW converters work in a similar way, so you can quickly get your bearings after starting them up for the first time. I can master even the most powerful ones within a day at most. Darktable works completely differently and also has an interface that makes getting started anything but easy. What’s more, the software’s incredibly complex.
For months I thought it would suddenly click and I’d understand everything. But that never happened. I understand the software better than I did to start with, but I’m still not getting really good results with it. Darktable’s hard to use, that much is certain. What I’m not sure about is how much the software can actually do once you’ve mastered it.
How modules work
Like many RAW converters, Darktable’s divided into administration and editing sections – which it calls Lighttable and Darkroom respectively. The latter’s particularly complicated.
All the parameters you adjust during image editing are organised into modules. There are over 60 modules in total, which you can organise into presets. You can modify pre-installed presets or create new ones. In the examples in this article, I’m using the pre-defined module collection for beginners.
Darktable processes the modules from bottom to top. For effective editing, you need to do this too. I didn’t know this for a long time, which meant no usable results.
Interface: not as bad as a year ago
First, I need to address yet another issue: the user interface. When I first started using Darktable over a year ago, I thought it was terrible. Everything was grey on grey, with such low contrast that you could hardly tell whether a module was enabled or not. Plus, you could only scroll through the modules if the mouse cursor was placed exactly on the narrow scrollbar. Otherwise, Darktable would change a randomly selected parameter. Menus and right-click context menus were practically non-existent, at least in my Mac version.
Darktable now recognises the concept of context menus, but still doesn’t make enough use of them. You can choose from several themes, some of which offer good contrast. YouTuber Andy Hutchinson has created his own CSS template, which distinguishes the individual modules more effectively. The scrolling issue’s also been resolved. You can now specify in the settings that, when you scroll, you actually want to scroll rather than change a setting. Maybe this option existed before, but now you get a notification about it when the program starts.
I still don’t think the user interface is particularly good – but it’s a lot better than a year ago. The changes are encouraging.
Scene-referred vs. display-referred
Darktable is currently at version 5.6. Version 3 saw a change in its base principle. Since then, Darktable’s operated on a scene-referred basis instead of a display-referred one. What does this mean?
The term originates from the film industry and means that the colour and brightness values of the actual scene are used as a reference – not those of the result on the screen. This means the editing process is independent of the output medium. At least, that’s how I understand it.
What I still don’t understand is how this is relevant to photography.
Misinformation, such as that found on darktable.info, causes confusion. It claims that a RAW converter usually forces the image data into the screen’s colour space and then works with the limited gamut. Nonsense. RAW converters access all the information provided by the image sensor. The data’s only reduced when you export as a JPEG.
Darktable.info isn’t the official Darktable website; it’s a community project. The user manual on the official site describes things differently and doesn’t sound wrong at least:
_In the real world, «pure black» does not really exist (there is always some light) and there is no limit to how bright things can be (so no «pure white» either). Scene-referred processing attempts to retain the physical properties of the scene for as long as possible by placing the RAW data on an unbounded linear scale and only compressing the data to the dynamic range of your display after image processing is complete. _
However, these considerations seem irrelevant when it comes to photography. A photo sensor does have an upper brightness limit – otherwise, overexposure wouldn’t exist. If the sensor’s completely overexposed, the brightness values of the affected pixels are all at their maximum. If this happens, nothing can be salvaged through post-processing, regardless of the RAW processing method.
To illustrate, here’s a completely overexposed photo. Essentially totally white, I can darken it in the RAW editor – then it’s a uniform grey. Either way, all the pixels look the same, with no image structure. It’s the same in both Darktable and Lightroom.
Why is the switch from display-referred to scene-referred settings even important? It stops certain older modules in Darktable working correctly so you can’t use them. This includes the white balance module. Confusingly, this module’s still included in the presets and has to stay enabled, but it shouldn’t be changed. Instead, I make white balance corrections in a new module called «colour calibration».
Plus, the switch introduces a number of new, hard-to-understand modules such as AgX or Sigmoid. These carry out tonal correction at the end of the editing process, but I haven’t been able to achieve good results with them so far.
Where are the examples?
What my internet research didn’t find were concrete examples of photos with before-and-after comparisons and information about the settings used. These would provide obvious, transparent arguments for Darktable’s capabilities. But they’re almost impossible to find. The entire – very comprehensive – user manual doesn’t contain a single example photo. I found this guide based on a single example photo. It’s quite good, but the example photo doesn’t have anything particularly tricky to correct. There’s no underexposure or overexposure, no harsh lighting, no image noise, no distorted geometry. A RAW converter that can correct this image has barely earned its swimming badge, let alone its lifeguard certificate.
My example of failure
So, let me walk you through the process using a difficult example.
The sun’s shining through the trees onto the ground, creating a high contrast between the brightest and darkest parts of the image. The aim is to balance these contrasts so the texture’s still clearly visible in the shadows, while the lighter areas aren’t overexposed.
The exposure’s preset to +0.7 EV by default – another Darktable quirk. More importantly, the correction process is fundamentally different from other RAW converters. The goal isn’t to neatly distribute the brightness values across the histogram. Underexposed and overexposed areas are explicitly permitted – only the main subject should have the correct brightness.
In an image like this, I don’t know what the main subject should be. I choose the bush in the centre of the image and the shadow it casts, so I brighten the image – setting the exposure to +2 stops.
Following the bottom-up approach, next up is the tone equaliser. Since the image is obviously too bright, I significantly darken the brightest areas and this is the result:
The brightness distribution looks okay now. However, the subtle colour gradations are missing. The image reminds me of a GIF with 256 colours. Or a JPEG that’s been processed more than its limited file size allows. I can adjust a number of masking sliders in the exposure module, but none of it changes this fundamental issue.
The colour balance RGB module allows you to increase the saturation in three different ways, with roughly the same effect each time. I previously adjusted the blue colour separately in the colour equaliser module because I wanted it to be darker and a bit more vibrant.
I haven’t touched the colour calibration – the module now responsible for white balance. I find the colours reasonably acceptable. But now there’s another mystery: the new module doesn’t have a green-magenta scale. This would be particularly useful for forests and similar subjects, since the camera’s automatic settings often mistakenly attempt to correct a green cast.
Sigmoid’s one of several modules used in scene-referred image editing to compress tonal values for the display medium. This happens at the end of the editing chain. You can only use one of these modules at a time to prevent them from interfering with each other. Sigmoid’s the default module for the beginner’s workflow, where you can adjust the contrast. A second slider determines whether the contrast’s applied more to the lighter areas or the shadows.
To be honest, I don’t get this module. My problem isn’t that the colour or brightness range is too wide – it seems to be too narrow. Logically, Sigmoid can’t fix this. Neither can the downstream local contrast module nor alternative tone control modules such as AgX. No matter what I adjust in these two modules, the photo doesn’t improve. In fact, it can sometimes get significantly worse.
Considering the difficulty, the result isn’t bad. But it’s not as good as in Lightroom. That’s clear to see even by looking at a cropped section.
This leaves me confused about the advantages of a scene-referred workflow. So far, it just seems complicated and unintuitive.
Show me how it’s done!
I suspect I’ve missed something fundamental. If you’re more proficient with Darktable than I am, please show me how it’s done. My sample photo’s available to download, along with the XMP file containing the editing information. If you manage to create a successful edit, please send me your XMP file – ideally with a brief description of what you did.
My interest in IT and writing landed me in tech journalism early on (2000). I want to know how we can use technology without being used. Outside of the office, I’m a keen musician who makes up for lacking talent with excessive enthusiasm.
Interesting facts about products, behind-the-scenes looks at manufacturers and deep-dives on interesting people.
Show all
