Devs - How Hard Would It Be To............?

Software RawTherapee
41

Learning a new program will always take some time. You’re not going to jump into C1 from rt/dt and be immediately proficient. Digital manipulation tools are inhernetly complex, as the problem is complex. I spend hundreds of hours learning photoshop’s many techniques and getting acquainted with all the various tools. If you are a frequenter of any other photography forums, you’ll find the same complaint about RawTherapee, ironically. To quote some dpreview users: “A GUI that only a PhD could love.”

If you start with darktable by trying to fiddle every knob and slider of every module, sure, you’ll be overwhelmed. However, the project makes recommendations of which modules to use, but we leave it up to the user to decide what works for them. If you spend a tiny amount of time with the manual and a youtube video, then you’d probably quickly learn that you don’t need to diddle every knob. For me, that is what is most disappointing about Andy’s initial post-- you’d think a person who makes instructional videos for people that go fairly in depth would be acquainted with doing a little bit of research, reading the manual, and experimenting a bit before drawing such rash conclusions. This discounts the fact that there are numerous incorrect statements in first post (as he is figuring out now by asking questions).

I’d say that with the scene referred workflow, the modules are less intertwined than they’ve ever been. These set of modules operate on specific parts of the image to give you precise control, and they try very hard not to step on the toes of other modules. If your perception is that it is a dense impenetrable mess and you don’t make any effort to counter that perception, then it’ll never change.

If you want to evaluate them, then yes, it can take some time. But, again, the project recommends modules to use and where to spend your time: darktable 4.2 user manual - process

I’d suggest you do not speak on something you clearly have zero understanding of. Again, you’ve absolutely failed at doing any type of research, the reasons why he forked are clear and you’ve completely whiffed on the reason.

We don’t need a poll, it is a commitment that the project made to its users.

Further, you don’t even see all the modules on first start, you see a select subset. The project is continuing to refine this subset, and the grouping will be updated for 4.4

I wonder in general: how long did it take you to learn RawTherapee? Was it the first day you installed it? How deep do you really go with the modules? How about the more technical modules like Wavelets? Isn’t a lot of what you’ve leveled here apply equally to RawTherapee, save the fact that you’ve put in the time to learn RawTherapee?

14 Likes
42

Andy, thanks for your library of teaching videos! They are a great learning resource and I have learned a lot over the years from watching them.

1 Like
43

No argument here. But my point is… once you have learned a program or two, most of the time you have a big head start learning a new one because the basic functions are the same. For instance, the interface of C1 and Lightroom are very similar and most of the functions are familiar and work much the same. RT’s interface is well delineated as to functions and is easy to navigate; some of the individual operations are obscure but they aren’t essential for basic editing and can be learned later at your leisure. ART is a little different and a little “easier.” DxO is so easy it’s kids’ play.

In comparison to all the above, dt is a real bear.

I’m not bashing dt, I’m just frustrated with it because it is a one-of-a-kind program that has so much power. All the efforts to help its usability are helping but nothing would be better than a good housecleaning and some streamlining.

44

But if you enable the beginner preset, you get:
image

Then turn on denoise if needed; lens correction, if needed. rotate the image by adjusting the slider, if needed. Adjust exposure using the slider, if needed. You can skip color calibration (maybe that could be removed from the menu), adjust vibrance and saturation to your liking in color balance rgb, click the pickers in filmic rgb, and enable local contrast if you feel you need more punch.

(Note that this screenshot is from pre-4.2, so the preset may have changed since then.)

Then, once you want more, open the modules with their corresponding icons. When you want even more, are fairly confident with the basic modules, select one of the presets modules: default or workflow: scene-referred. When you are even more advanced, and know what you want to use, create your own.

3 Likes
45

You should reread your first post; this statement is at odds with it.

Not true, you just haven’t encountered something like it. This workflow was certainly not invented by darktable devs, and looks very familiar if you’ve edited video or done some 3D animation.

As I said above, the project is constantly refining what a first time user sees.

2 Likes
46

Yup, and most of RT’s workflow already follows this - linear RGB up until the tone curve phase. RT is just not as pedantic/obsessive about it. I can’t speak to local adjustments in this regard, that’s a feature I haven’t done much with.

I do think the sigmoid module from dt is excellent and a candidate for somehow shoehorning into the tone curve UI in some way, but I’m not sure how. That said, many of the benefits that darktable users are seeing from it are already in RT, and in fact were inspired by RT’s implementation, it’s just a few corner cases that I think RT could handle a bit better. That said, most of those corner cases are ones where you probably want to flip on some sort of local tonemapping.

There are a few cases where image operation reordering would be useful, but it’s mostly a nice-to-have. Multiple module instances would inherently come along with image operation reordering. That is one case where the local adjustments UI probably could use some cleanup, because it effectively gives you multiple module instance and module reordering - just at (unfortunately) a highly fixed location within the RT pipeline.

One thing that DT definitely does better is output color profile management - it’s almost always tied more to the output format than to the current image, and I don’t like the fact that it’s per-image in RT - Input/output color profile support has a clunky UI · Issue #6644 · Beep6581/RawTherapee · GitHub is part of this, but some of the things I’d like to see (or attempt to implement) go a bit beyond this - e.g. selecting the output color profile in the export dialog and applying it to all images.

3 Likes
47

I’m not a wordsmith like you and am not very elegant in the way I phrase things, but dammit, could you drop the attack dog posture for a few minutes and address the essence what I’m trying to say?

I said, paraphrased in several places across several posts, that dt has impressive power and many unique features, but also has serious issues that make it extremely difficult for users to learn and for coders to upgrade and maintain. Nothing new, you know it’s true, and it’s a subject of many, many posts in this forum.

The question of why has been posited often: the dogged obsession with 100% compatibility with all past edits has caused dt to become unnecessarily complex, bloated, and tainted by compromises and workarounds needed to accommodate it. And it will just get worse as time goes on.

AND THE SAD PART IS THAT THERE IS A POTENTIAL SOLUTION BUT IT’S OFF THE TABLE: bite the bullet and break the chain… jettison the obsolete, keep the best, and release a new streamlined version. Past edits could still be maintained with an older version of dt. What would be so bad about that? I bet most people would like this.

If Aurelien could do it as a one man operation (R&Darktable), just think what dt’s team could do.

Mica, you have been and still are making great contributions to dt. You are smart and you are dedicated, no question. Please, don’t be so quick to (mis)judge and belittle people you disagree with.

A challenge to you. Since I don’t know of any, maybe you can tell me: What other program(s) are like dt?

dt’s workflow is not as straightforward and predictable as you imply, due to all the issues previously mentioned. It COULD be much better if dt were streamlined.

Waiting for dt gold - 2023 edition.

2 Likes
48

Do you think RawTherapee is any better? Because in terms of code quality and clear UI/UX design principles, my humble opinion is that darktable is leagues ahead of RawTherapee.

Again, do you think RawTherapee is any better? Have you ever tried using Retinex (does that name even tell you what it does?) or Wavelet levels? And can you explain to me in which order you should use the different modules of RT so that you get consistent and reproducible responses to your edits?

2 Likes
49

I’m pretty sure there are a lot of people who appreciate that current dt can open their old sidecars… LR manages such with their processing baselines; you just can’t ignore folks’ years of processed images.

Keep in mind that dt has recently undergone a sea change in their pixel pipe, one that significantly improved rendition outcomes. That the UI survived that change is quite significant…

However, sometimes it is just right to start anew. You might want to consider vkdt, whose current UI you’ll absolutely hate. But IMHO the underlying processing architecture is really what’s needed to provide a cleaner foundation for any UI. vkdt raises the bar for entry, however, a decent GPU is required, but doggonit, we’re processing images here and that device was purpose-designed for the task…

5 Likes
50

I don’t feel that quoting what you’ve already said and replying to it is “an attack dog posture.” In one post you say “I’m not bashing dt” and in another you say “dense incomprehensible mess” and “Is this really necessary? Is this fun yet? No wonder Aurelien left” and “Is dt this convoluted…”. So your statements seem at odds with each other.

If I’m reading this, and some of your other comments, correctly, what you’re trying to say is “darktable has too many modules” Is that correct? If you mean something else or anything else in addition to that, please explain.

If this is also meant for the “too many modules” type of comment, then you’re incorrect. Old modules don’t change codewise and don’t require too much maintenance. Ripping all of it would likely be a lot more work than leaving them alone.

Where has this question been posited? Please provide links.

As a fairly long term user of darktable, I’d be very pissed off if my edits suddenly stopped working. My library of images is likely quite small compared to other users, but being able to open and compare my old edits is extremely useful to my creative process. Having to run an older parallel version to view or change older edits would be a significant hardship on me, and I’d suspect on many others as well.

It is unclear what you think Aurelien has removed in his fork. And you don’t appear to be correct in your assumptions. I consider comments like this to be “bashing” dt.

A judgement would be my opinion, but this isn’t my opinion, its a fact. darktable derived the scene referred workflow from 3D animation and video editing software. That is where the controls for modules like filmic and color balance rgb come from. It is different from other photo editing applications, but, again, the dt team did not invent these tools by themselves.

Isn’t the reason why ART forked from RT so that a developer could remove the older modules from RT and replace them with newer ideas without having to worry about breaking legacy edits? And isn’t that the reason why some modules in RT continue to hang around?

I can not understand how you overlook that irony.

I would also consider this type of hyperbole as “bashing” dt. It seems unnecessary.

I don’t think the UI/UX of darktable is perfecct, it is far from it. But I also don’t think it is as bad as you state. I think for a long time darktable has provided tools that operate with more nuance than commercial alternatives: that’s why the Contrast Equalizer module had three tabs, where as Lightroom’s Clarity only has one slider. The same idea of nuanced editing extends now into a bunch of modules: filmic, diffuse or sharpen, color balance rgb, tone equalizer’s masking. If you’re the kind of user who is happy with one Clarity slider, then you’ll never be happy with darktable; the ethos is different. However, if you wanted to effect edge sharpness apart from luma contrast, you’re never going to be able to do that with the Clarity slider-- those functions are tied together.

As already mentioned, the spiritual successor to darktable is called vkdt, by the original author of darktable, @hanatos I don’t think vkdt is ready for primetime as it doesn’t have enough features yet, but hopefully the many lessons learned from darktable can be implemented here. Some already are: the GUI loads from a text file, the processing pipeline and GUI are completely seperate, the processing pipeline is nodal instead of linear.

I also hope that with some community participation, we can balance adding new features with application design clarity. That remains to be seen though.

8 Likes
52

Who said anything about complex modules? If a module needs 2 buttons, or 5 buttons, or 25 buttons and 15 sliders to perform its functions, fine, you learn how it works if you want to use it. That’s not the issue. The point is that even a person who has years of experience in photo developing can’t open dt for the first time and expect to do some basic editing in 15 minutes and a lot of the more advanced editing in less than an hour like he could with many, maybe most other programs out there including the big commercial ones. But there’s no technical reason dt couldn’t be that way, and it wouldn’t take a total rewrite to achieve it.

The design of the interface isn’t the problem, especially with all the changes that have been made to group the functions more helpfully. The problem comes when you need to reach beyond the basic modules to do more advanced editing. There is so much duplication or near-duplication (and that in itself is confounding) of function across multiple modules, some of which is interactive with unintended consequences, some of which can impose undesirable constraints downstream. You need to refer to a roadmap until you memorize enough of the foibles to avoid them.

AP ruffled a lot of feathers here when he tried to get dt’s path changed. His current effort, Ansel, is “…a darktable fork with 22229 lines of bad code, bugs and ill-designed features removed.” AP’s demeanor is tough to deal with but I think his vision of what is and isn’t “good” software is much clearer than what is present in dt’s arena.

The old salts around here that have memorized everything place all the blame for poor dt user experience mainly on user laziness. Hmmmm.

dt is an incredible piece of work that could be much more accessible if it were made more user-friendly. It’s a shame that the powers-that-be choose to keep it cloistered.

53

Look, I also prefer to have the old modules around, and being able to re-open a 10-year-old edit. The dt devs are trying to manage backward compatibility and also steering users to what they consider to be the best modules nowadays, by deprecating and hiding (but not removing) modules. They also providing module group presets with a recommended set of modules, starting with the rather bare-bones ‘beginner’ subset. There is never a single solution that satisfies everyone. Some tools take the ‘opinionated’ path. Filmulator is a great example of a unique application, providing a few select tools and rigid pipeline that work together very well and lead to an easy-to-use software. Then there is ART, which took RT’s codebase and started, at first, streamlining it, then turned it into a great alternative tool with its own developments.
We’ll see what Aurélien will achieve, it’s too early to judge. He certainly takes the opinionated path, but it seems to me that his opinion is that everyone but him is a ‘coding monkey’ (his words). Long gone are the days of AP writing acknowledgements like here: Filmic, darktable and the quest of the HDR tone mapping - Aurélien PIERRE Engineering.

4 Likes
54

And I think THIS is the root cause of most of dt’s usability issues.

My complaint has never been with dt’s UI design per se – I think it’s very well done – but with the fact that dt’s user experience is degraded due to issues with the modules themselves: the sheer number of them, the presence of obsoleted modules, the duplication and near duplication of module functions, the unintended interaction of some modules. Which of sometimes multiple choices is the best one, and why, and when, and are there unforeseen downstream consequences? In many editing situations a good way forward just isn’t clear.

Editing is already difficult enough without having to fight the program itself! This kind of poor user experience would get heads rolling in any industry.

To answer your question about RT, yes and no. RT is much less complex than dt and doesn’t have dt’s raw power and breadth of capabilities but it does have some superior features that are as good as and sometimes even better than anything else out “there.”

What does this have to do with the issues I raised? But yep, I have used Retinex and Wavelets. I started using RT what, more than 10 years ago at verion 2.41 when Gabor still ran it. While I’m not a technical expert with any program I can usually work or kluge my way far enough through most of their features to get a decent result (within my capabilities). Including Retinex and Wavelets.

55

Long time dt users like yourself have memorized everything and have enough experience to instinctively make good choices in many cases. But what about the rest of us? We’re left with a very tough row to hoe.

What if Microsoft would have made the same kind of decision about keeping all programs compatible with Windows XP? Remember that kluge add-on 64-bit extension they made to it? It kinda worked but not very well… Windows needed a new direction. What if Linus didn’t act on his idea for a sort of slimmed-down UNIX?

The compatibility feature is the root cause of many, maybe most, of the usability and performance problems that dt has.

What would be wrong with making a clear cutoff, say there’s a new release on June 1, 2023, where backwards compatibility with version 3.0 and earlier is no longer guaranteed? You could still keep 4.02 or whatever installed to handle all those old edits at the cost of only a little hard drive space. But dt and its users would benefit in every way imaginable, and more and more as time goes on.

I know this is all a pipe dream, that it’s not going to happen any time soon, but, well, just sayin’…

56

It looks promising but unfortunately it won’t be available to us Windoze people for a while.

57

As long as the developers make that sacrifice, and take upon themselves the burden of maintaining the code, I don’t think it’s an issue for users.

With the beginner module groups, one gets:

image
image1000×381 27.5 KB

I don’t think that’s overwhelming.

And this is my module group setup. It has all the tools I use often: corrections for noise, chromatic aberrations, rotation and perspective; tone and contrast manipulation; colour adjustments. LUT 3D I hardly ever use, I could remove it, in fact (since I still have the search bar):
image

To sum it up: darktable is not perfect. No tool is, ever, since we, users, are different. I’ll now really give this thread back to Andy and his suggestions for RawTherapee (which was my first raw developer; my first pp2 files are from November 2009).

3 Likes
58

You keep saying that but it’s invalid. As long as you change your workflow to scene-referred as it’s explained in the manual, you will not see any legacy modules, they will be hidden from sight.

1 Like
59

They do! To a ludicrous degree, actually, including custom shims that are shipped with Windows, and activate special workarounds for specific programs.

Linux doesn’t, and in fact Linux software is not very compatible at all across distributions or versions.

3 Likes
60

Aw, come on, Wavelets is really handy, as I’ve demonstrated many times. The idea of wavelet image separation and processing has been around for donkeys years.

It’s like Photoshop frequency separation on STEROIDS but using raw data rather than rasterized pixels.

But I’ll admit ‘wavelet’ as a name is not real intuitive for those who have not experienced the full history of digitized imagery - yep, that means I’m an old fart!

As for Retinex, well, I don’t think I’d miss it if it wasn’t there - can’t remember the last time I used it.

CA&L module makes me smile - ages trying to do something that takes seconds in Photoshop or CameraRaw and smart objects, though I do understand that not everyone has Photoshop for whatever reason.

1 Like
61

So… Too many modules. This has been addressed several times by several people in this thread. For some reason you haven’t replied to any of those comments, but you choose to quote other things then repeat yourself. Its very frustrating because the conversation does not advance.

You insist to get rid of the legacy support, accuse long time users of “memorizing everything” but can’t critically look at the application you use and see exactly the same thing. All irony is lost on you, or at least you will not acknowledge it.

You seem to have no clue what Ansel sets out to do or what caused the fork, but you continually talk about it as if what you say is facts.

You’ve skipped answering every critical question posed to you. What’s the point if the conversation never advances beyond your own slim vision?