Looks good.
Some basic good old film terms explained here:
http://www.sprawls.org/ppmi2/FILMCON/
Might be of interest?
Have fun!
Claes in Lund, Sweden
I have questions regarding the shape of the filmic ‘S’ curve … or more directly about the ‘toe’ of the curve.
- When the toe lifts away from the lowest position on the graph, what is the graph telling me? … and secondly:
- When the toe forms into a ‘hook’ or starts to form a hook shape, again what message am I getting?
It means the target black luminance on the display tab has been lifted above zero. This is now done by default, because an 8-bit JPEG can only show 12.7 EV of dynamic range, and so if we set the look curve below 0.0152, any values below that point will all get rounded to zero, and you’ll no longer be able to discern any details down there.
The hook will be highlighted with an orange indicator, and it means you have set the contrast too steep, the latitude too wide, or the shadows/highlights balance too far on the shadow side. In these cases, filmic has trouble to produce a monotonic spline, and so you will get gradient reversals in your image. To correct this, address one of the 3 points above, or reduce the target black luminance on the display tab (if you don’t mind the black clippping that may result in your final JPEG) and these actions will ease off the constraints on the spline.
1 Like
There’s a bit of fundamental-ness about any tone curve that’s worth understanding, that of the x-goezinta → y-comesouta, the so-called “transfer function”. With that, any of the tone curves can be understood to their effect. Of note is that the “identity relationship” of a tone curve is the straight line from lowest-leftest to highest-rightesst: if you take any x value, follow it straight up to this line, then straight left to the y value, they’ll be the same. It’s when you make the line depart from this diagonal that you actually get a difference…
So, I’m going to change the question slightly, ‘when the toe lifts away from the diagonal’. With that, the lowest values will be increased, making them brighter. Of note is that the lowest adjacent tones will have more difference between them making them more contrasty.
Same sort of minor re-definition, ’ when the toe is pulled below the diagonal’, in this case the lowest values will be decreased. The slope of the curve for those adjacent low values will be smaller, and thus contrast will be decreased.
The filmic toe was particularly noted in Duiker’s original presentation, as providing “crisper blacks”. Not a very precise description, but the specific behavior of the curve from 0,0 was to run below the diagonal initially, then increase the slope at some point to climb back to the diagonal. Well, this curve shape starts the lowest values at a very low level, then introduces contrast as the slope is increased (note the relationship of slope to contrast). The contrast at the low end is what makes the “black” tones “crisp”.
In plots of filmic curves, these toe characteristics can be hard to see. You need to be able to zoom in on the left side of the curve to see the “real” toe. John Hable has a good blog post on the whole topic, including a couple of curve plots that show the toe:
2 Likes
I think that probably 99% of film photographers, even professional film photographers, would find that of pretty much no help to very little help at all. That sounds like the stuff that a film engineer or film scientist might be very well versed in.
IMO, jillesvangurp has made an excellent point. I won’t comment on his ideas for a solution though because I don’t know. The idea of an expert mode and normal photographer mode, I think, would be helpful though.
And also I have seen mentions in various places about how some dt modules don’t work well with filmic or shouldn’t even be used. When using filmic, clearly marking those modules, hiding them, or something else would reduce the complexity and confusion, IMO.
I hope not!
I like to have everything available to get the look for my picture that I want even if they are not supposed to work nicely together
I sometimes use a chisel to open a paint can, it’s not recommended but it works if used carefully
darktable isn’t lightroom without subscription.
If you want to hide the tools you won’t use - darktable has a solution for that. But you need to do it yourself.
Those who need to be spoon-fed are not the target audience of darktable
1 Like
That is a bit rough… One must admit that dt has taken a very engineer/scientist path these last few years, and even if there is no doubt that the algorithms quality has vastly improved, the suggestions on GUI improvements should not be discarded with arguments like that.
I understand that dt devs have no will to make it a mainstream piece of software but there is nevertheless room for improvements for the non-engineer users, even without sacrificing its underlying power. And some of the above suggestions are very valid. My 2 cts.
1 Like
Maybe rough, but there isn’t a one fits all approach if it comes to UX. A tool for experts needs a different UI than a tool for an user searching for a foolproof workflow.
1 Like
- Everyone thinks he/she is the epitoma of the “average user”, while they really have no idea. I ran a survey, the darktable average user has at least a master degree. I’m not saying we should disregard the others, but that’s what it is.
- There are all kinds of professionals. Some did the whole analog printing process themselves, some were happy to let it to lab techs. So we need to adapt explanations to every level of expertise. I know lab techs who learned filmic in an afternoon because it is exactly what they are used to.
- A film scientist/engineers designs film. What is presented here is general culture of film for people using them. I have learned in high school that light is “made” of photons. That doesn’t make me a quantum physicist. I have only general culture in quantum physics.
There is no normal photographer.
Everyone here needs to understand that the issue is not expert GUI vs. amateur GUI.
GUI only control pixel processing algorithms. Algorithms need to achieve specific tasks while adapting to all kinds of inputs. The main issue is, I, the developer, have literally no idea what you feed to my algo.
It can be a DSLR raw, it can be a scanned negative, it can be a TIFF processed somewhere else. Black can be corrected to RGB = 0, or not. White can be anchored to 1 or more. Plus I don’t know if you exposed to protect the whites, or to anchor the middle-grey, or you ETTR. So middle grey can be anchored at 18% or anywhere else. Metadata contain lots of crap but not this.
Because I have no way to know, all these parameters need to be exposed to the user, because only him/her knows the whole life cycle of the picture.
That’s why the GUI is crowded.
Any other software gets away with that because they use a display-referred workflow, in which we know white = 100%, black = 0%, middle-grey = 18% because it is designed to make that happen. That is achieved by applying non-linear mapping transforms straight after the input color profile, very early in the pipe. But that comes at a price : alpha blending is broken in display-referred workflows (that’s actually the only reason why the cinema, 3D imaging and CGI industries don’t use it), and any optical filter will fail in non-linear display-referred, because optics laws are violated if you re-encode light emissions non-linearly.
I’m afraid it has nothing to do with target audience. If you want pixel-correctness, you want scene-referred workflow, and you can’t assume fixed values on black, white, grey, but no GUI design magic will spare you the trouble of setting them, because the software cannot know the previous life cycle of the image you feed it. These things are not written in metadata.
It’s not a “geeky” choice of design. First, there is pixel-correctness. Then, there are algos to achieve it. Finally, there are GUIs to control such algos. And all of them are complex because what they do is complex. For 20 years, that complexity has been hidden for the sake of simplicity, but you pay a high price for that : bad colors and difficult processing in HDR-like situations.
I have been on the darktable’s mailing list since like 2012. Many people have complained about issues with colors shifts and over-saturation, usually blaming input color profiles, while the real culprit was the base curve coming first in the pipe. To these people, I want to say there is no solution in a display-referred workflow, the workflow is the issue.
Scene-referred brings solutions to problems that can’t be solved in display-referred. I never said it was easier. I never said it was nicer. I didn’t choose it because it is more sciency/geeky. I lost dozens of editing hours trying to bring back backlit pictures in darktable 2.2 and 2.4. I couldn’t. Bringing back contrast made me lost color saturation, or preserving saturation made me lost shadows. Scene-referred makes it possible to edit HDR scenes while getting a believable result. Not easier.
I’m all for simple GUIs and easy things. Unfortunately, I have no realistic actionable idea to make this happen without forfeiting pixel-correctness. So, be my gest. But vague stuff like “editing should be easy” are not actionable solutions, I’m afraid.
23 Likes
I think there are ways for GUI to be simplified without sacrificing anything.
When editing my images, quite often I only adjust few sliders:
- exposure from exposure module
- white relative exposure on filmic
- black relative exposure on filmic
- the strength slider on denoise - wavelet auto
- output saturation on color balance
Yes I do use reconstruct and look tab on filmic too, as well as other modules and more tuning, but not quite as often as the above. I appreciate the power the current GUI gives me, but I fully understood people who appreciate to see less options. I mean, how many people would need to adjust the display option in filmic module. My point is that there are small set of actions to produce acceptable edits.
So I’d say for photographers who are not into learning to use the tool properly, they can be presented with simplified options. Maybe a global setting in DT for the “dumb” version and the “advanced” version. When in the dumbed down version, all modules should display only set of parameters that are more commonly modified.
I know it will be a big task to classify which parameters are basic and which are advanced, but I think it can be done.
Anyway thanks for your work and dedication on DT. I do appreciate all the “advanced” stuff and your videos which make me keep learning.
1 Like
I don’t share your opinion. I think Darktable is quite different from other software and there’s not a 30 minutes transition from Lightroom to Darktable no matter how many features you hide.
Everybody tries to figure it out on their own when they first try Darktable but in the end they are destined to read the manual either by going there themselves or when urged by the people on this forum, twitter, facebook etc.
The thing that you are asking is basically to hide features that aren’t even visible b/c they are not placed in the first tab in far example filmic.
What I would suggest and what would actually help is to make a nice onboarding screen. Like Blender, Darktable could have an onboarding screen where you could setup some default behaviors like setting up display or scene referred workflow by default. There could be a link to the documentation with suggestion of places to start reading. There could be an introductory message to read. Maybe “Tip of the day” (more like tip of the session) where we would show a useful tip to the user in a short one sentence. There is a lot more that could be done in onboarding.
There shouldn’t be a link to this forum tho, we would be swamped by angry users who didn’t read the manual xD Ppl need time to adjust and accept that they don’t know anything about image editing.
Also the current help tool in Darktable links to non existing documentation on GitLab (404). At least for filmic. That should be fixed.
I think this is why lightroom exists. If that is your attitude as a user, lightroom is the right choice.
4 Likes
Lol, I don’t even use Lightroom, ever. (Okay I admit I helped my wife when she had problem uploading her collection to her Lightroom cloud, but that’s all about it.) I have no idea how the Lightroom editing experience is, Darktable is my first and only photo editing tool.
But why we always refer to Lightroom when given this problem? It really is not DT vs LR problem, it is about how to improve DT usability and user experience.
Yes you quoted my sentence where I said people who don’t want to learn. I don’t meant to say literally like that. Even for Lightroom, people need to learn too. It is more about different degrees people invested in learning the tool, or understanding the under the hood (or how it relates to light and physics).
Few of you said the best usability is to give the user all the knobs needed to control the car / organ / stradivarius. But many people (which may not be the current DT user base so it won’t show up in Aurelien’s survey), don’t need all the knobs. Because maybe 50% of the knobs is sufficient for them to produce good images. People will eventually learn the other 50% or whatever they need to learn to do their job.
Aren’t those people deserve to get a great tool too if it is available?
I really appreciate the improvement in scene referred workflow in 3.2 + filmic 4. I think that is big improvement and make me understand the tool better. So that is saying that the previous GUI can be improved. I’m saying it is possible to improve the GUI even further.
If we can get past the point where we are debating whether or not the GUI should be improved, maybe we can be more focused on the how.
2 Likes
Wait for v3.4 and those should be fixed.
1 Like
But then, if someone doesn’t want to learn the tool, why do they bother with dt at all ? Wouldn’t the camera jpeg combined with a simple graphics editor be a better choice (for cropping and some basic adjustments) ? Especially given how much time camera makers put in to automate as much as possible (including face detection, auto framing, auto HDR etc.).
Yes, ppl need time to adjust to a new tool. But, more importantly, they have to realise and accept that they need to spend time and effort to learn that tool (that includes spending time at least browsing the manual). And that might be the real problem, which no programmer can solve.
In at least some professional environments, introduction of a new program means training sessions for those that are going to use that program. (Those sessions have to be paid for, btw, usually by the employer). Why should it be different for a complex tool like darktable? Especially for those that plan on using dt in a professional context, they should seat aside time for learning the program, as they probably did for whatever they used before switching to dt.
2 Likes
I’m no longer a Darktable user primarily but when testing recent builds it seems to me that filmic is now basically a base curve replacement. (wait) In that you can most often leave it alone. Just adjust exposure and go on doing the other things you need to with other modules.
This, if true, would mean that beginners shouldn’t even look at filmic and instead focus on other modules? What do you think?
1 Like
Because adobe has spent a lot of time and money abstracting the complicated into single slider solutions. They have done a pretty good job of it too. There is no shame in choosing the right tool for you.
You have to say what you mean to communicate effectively.
The learning curve is much less steep (and the peak not nearly as high) and the learning material is more abundant.
So the user should ignore the other sliders until they’re ready to experiment. In an earlier post you suggested a beginner and advanced GUI, and that is almost always a non starter.
Those users who are not willing to invest in learning should uslurp time from developers in order to cater to their unwillingness.
This line of thinking often stems from the notion that darktbale needs more users or is trying to attract users in the same way a commercial product must. This simply isn’t true, and if darktable never added another user, it’d still carry forward as it is. The developers are making tools for themselves, and while there is wiggle room to allow for improvement and input from others who can’t or won’t write the actual code, there is not room for a complete paradigm shift (towards catering to the user who is unwilling to invest in learning).
I don’t think there is anyone who thinks the GUI is perfect and finished. But how it can be improved without wholesale sacrificing flexability and power for ease of use is a lot harder than most think.
2 Likes
I think we have to differentiate between ease of use for the experienced user, and ease of use for the novice.
With the current interface, you can do a lot to make it easy to use for your needs, using styles and presets, and by hiding the modules for which you have no use, and putting the onse you use all the time in the favourites tab.
But this of course requires you to understand a minimum of how darktable works.
Ease of use for novice users usually comes down to “hide what they don’t need”. But why would they not need a tool that’s available (and thus supopsedly has/had a use)?
Deprecated modules is another kmatter. Those should, I think, be marked some way in the interface. But that’s a one-off operation. (Given the way darktable works, removing them risks losing old edits, so is a non-starter).