Hello.
I consider Darktable default import settings are unusable both for tonal curve and sigmoid modes - the resulting images neither mimic in-camera jpeg nor the real scene captured. Any image in Darktable should be altered from the defaults to obtain acceptable results. I’ve made a bug report some time ago and there have been a discussion with no result.
So there is a question to the community - which settings do you use at import?
Here is my latest universal style. For me, the results are pretty close to a scene, viewed by naked eyes. The built-in jpeg is pretty close too. A7_filmic.dtstyle (3.3 KB)
Sharpness and noise reduction are optional. Color balance RGB can be altered to decrease saturation of highlights say to -25.
P.S. The problem of extreme highlights still exists for me - I guess Sony A7M4 in-camera jpeg highlights are better. One can see the built-in jpeg and compare. The purpose of -2EV exposure compensation was to preserve the highlights. DSC06355.ARW (34.4 MB)
P.P.S. I’m still sure the program should provide acceptable (at least) results out-of-the-box.
I use one of two simple styles and many images don’t need fiddling (except for denoising and correcting perspective and CA, which my compact camera produces in spades). However, it is not my purpose to imitate the camera look.
If that is what you want, try RawTherapee or ART, the curves tool in those has an auto-match camera curve button.
In darktable, you could also try using base curve instead of filmic rgb or sigmoid.
Everyone has a different workflow and different needs. I very rarely take “snapshots” or larger series of photos that I want to keep (e.g., family celebrations). For me, photography is a hobby to take “aesthetic” or “special” pictures (I don’t want to call it “art”).
That’s why I look at the RAW file with as few activated modules as possible and then decide which way I want to go. If I start with too many presets, the “look” is already too fixed.
I generally don’t look at the camera JPEGs (at least at the beginning) so as not to be influenced by them. However, if I ever require a larger number of images that I don’t want to develop individually, I can always fall back on them, as I shoot RAW and JPEG.
Who is to decide what is an “acceptable” result and, above all, according to what criteria?
Every camera has its own look. In addition, my Nikon alone offers six different look options that I can choose from.
There are also countless analog film looks that many people appreciate.
And what about different scenes? Should you have the same look for every scene?
Do you really need such strong contrasts in a scene like this, so that you lose all detail in shadows and highlights, and then have to struggle to get them back afterwards?
Why can’t I decide from the start what I want to do with my scene instead of having to follow the taste of whoever thinks he knows what is an acceptable default look for everyone?
The fact that there exist some difficult cases, f.i. high dynamic range scenes, is not connected with the fact that the majority of images can be processed with the same settings and provide appropriate finish results. That is logical - a user imports some images and most of them are good without the need for editing. That’s how all the other programs do - OS or commercial ones.
The samples with deep shadows are not relevant here - one can obtain required results from any initial settings, if he knows the program. They simply don’t matter. But if the initial settings suit most of the cases, it would require much less efforts from a user to process his images.
I’m pretty sure that there are some “universal” settings that can provide good results. At least I’ve used the profile above for great amount of Sony A99 images, several thousand Sony A7M4 images and some Nikon camera images and got the appropriate results out of the box (only exposure compensation changed). Of course, some images require editing, but not all of them.
As I’ve been using dt from it’s birth :), I’ve made color matrices and base curves for my former cameras by myself and used them. Now I use my own style which satisfies me in most of the cases. But I’m talking not on myself here. The fact the program produces dull, washy images out of the box, that look neither “real” neither as in camera jpegs is not good.
If you set the input profile to linear in most commercial software you get the expected dull muddy image similar to DT. Then you have to add lots of saturation a tone curve or two etc etc. There is little “magic” in all those other software packages they just do that out of the gate. I think you have to compare DT with a base curve to other software and then if it’s still not close then in DT would need to add more punch or as you have done create a custom base curve.
I think DT gives you lots of options to take either approach and to customize it which ever way you go. Personally I don’t bother with styles I don’t find they help. I still end up tweaking things so I look at my images. pick one edit as I like then paste the settings to similar images. Then edit a new key image when required and repeat… I find this gives a better match than trying to use a global style… But that is only my preference. Others might use a lot of styles???
Then please provide those settings. If they’re as good as you say, then they’ll be adopted by the project.
Personally I like seeing the image with minimal processing so I get an idea of what data I have to work with. Then I have a style that I apply that sets things that I generally don’t change. Then I adjust exposure and filmic. The image is generally acceptable at that point and I can begin local edits.
I have to say that I’m perfectly happy with the default sigmoid settings.
I do have custom matrix input profiles for a couple of my cameras, but the differences are very subtle.
To be fair. the fact that dt does not give a ‘ready to go’ image was a challenge for me when I first used darktable, a couple of years back, but I’m actually grateful that it forced me to learn more about what settings actually do what, in terms of image appearance.
I think the majority of darktable’s developers and keen users prefer the approach of starting with an image as close as possible to the actual raw file. And I find that a perfectly valid approach.
That’s your opinion, and a completely valid one. But the fact holds that darktable is developed by it’s users - and one can’t really expect the handful of users who actually put in hours on end of coding to make changes that they believe are counterproductive.
To me the beauty of darktable is that it does not simply copy the out of camera jpg. If that is the look I want then I would use the out of camera jpg or use the editing software supplied from the cameras manufacturer as that will produce the same look. Also RawTherapee comes close to the out of camera jpg look so I would choose that program over darktable.
I have a picture style which I apply to an image when I open it in the darkroom. It is simply denoise (profiled), color calibration module with WB set to as shot in camera, and initial sharpening instances using the diffuse or sharpen module. However, I have switched from using filmic as my default tone mapper to using sigmoid as it gives me a more pleasing starting point straight out of the box.
I feel you have to come up with some picture styles that give you an initial look that you are happy with as a starting point. But it seems you have already done that. There is no need for DT to provide an acceptable result straight out of the box.
As for highlights, I set my exposure slider to preserve the highlights and then I can use numerous options including tone equalizer, shadow and highlights, color balance rgb and other options to recover the shadows. If I sacrifice my highlights when capturing the image in the camera or when setting the initial exposure slider it becomes difficult if not impossible to recover later.
Good luck with your editing. It might be worth trying RawTherapee and seeing if you prefer the results obtained from that program.
I thought I’d put in a 5 cent contribution here while procrastinating on the Christmas Day sermon for tomorrow (Australian Eastern Standard Time).
Although I have recently added an X-touch mini to enhance my editing experience, I am of an efficient use of time (ok, lazy!!!) bent - I have a group of presets that auto-apply on import. This is partially so that I have the ability to compare, in culling mode, the particular captures easily without being distracted by under-exposure or poor tone-mapping. Once I am down to the few images that I wish to focus on (boom-tish!) I may sometimes throw away all the auto-applied presets and start with a clean sheet.
My auto-applied presets applied when I import from my Nikon DSLRs (but not from other sources like phone capture) are just 4:
exposure - auto mode, target level -3EV - a preset I call Martin-3
sigmoid - a preset called Martin - contrast 1.5, 0 skew
denoise (profiled) - non-local means auto mode
sharpen - preset called 04 Martin’s medium sharpen - radius 4, amount 1, threshold 0.5
It may not even be valid if you hold to using the word real… the real image is captured light and then its a green blur that ultimately becomes what it does from the math is applied to it so holding up the jpg as the ground truth is not the best reference. It is what you see on the back of the camera so I guess you can have that bias but then the camera processed image will be the closest thing to what you remember you saw in the viewfinder or screen on the camera so why waste time trying to recreate what you have…
If that is the goal then using darktable chart and matched jpg and raw images of a colorpicker will get a style that has a lut and tone curve and offers a pretty good match. Also there are basecurve tools that can be used to make custom basecurves and you could even put it back in the legacy pipe order if that looked more like a jpg but DT moved on to offer something a little different and so it just requires a bit of a different approach and depends on your priorities.
Yeah, and I don’t understand what’s so objectionable about having it turned on by default when you a load an image for editing.
The “How am I supposed to decide how I’d like to process my images if they’re not a washed out grey mess when I open them” argument is a truly bizarre one. Would it be better if the white balance wasn’t turned on by default too? Why should your artistic vision be hampered by what the dumb camera thought was appropriate?
Warmer and cooler tones are but a click away, as are contrast adjustments…
It’s easy for an experienced user to configure anything and everything to their exact requirements, so what’s the harm in making the defaults a bit more beginner friendly?
More users = more disagreeable opinions I guess, but please don’t take any of my opinions above as “complaints” per se - I’m extremely grateful for all the developers’ efforts and darktable is the best tool of it’s kind I’ve ever used. I’ve found a workflow that’s perfect for me. The unfortunate thing is that in spite of that I find it very difficult to recommend to others
Just to be clear, I do (these days) have darktable setup with the sigmoid option in the ‘pixel workflow’ setting or whatever it’s called in the preferences.
So sigmoid is on and the images (to me) look pretty good by default, out of the box.
I use profiled denoise and lens correction, and otherwise like the sigmoid defaults, with skew -0.2. I crop and adjust exposure.
Since no camera has the same spectral response as the human eye, all images require color correction and there is no way to automate it. I prefer to do this on a “plain vanilla” image before I do anything further to it, it is a distraction to deal with the interference of modules that otherwise change colors.
Then I apply the rest of the modules as needed.
If you want “out of the box” images automatically adjusted with minimal work, Darktable is not a good match for your purposes.
After setting exposure I find it looks quite good out of the box .
Almost all my images get exposure , tone equalizer , local contrast , two diffuse instances…
And denoise / chromatic abberations / lens correction , but only when opening the raw file directly. Often the raw is proprocessed by DxO Photo lab so then I skip those.
But everytime I make a style or I copy & paste a bunch of settings to a whole load of files , the first thing I end up doing is disabling almost all modules. Because I work with applying it module for module. I have an order in which I apply the modules , judge the effect and then move on. So in the end I skip any kind of default style , because I end up disabling the module and later re-enabling it again.
