To reiterate what Todd already said: this is just an experiment. We don’t know if or when it will be included in darktable, and if it does it may be in a different form and the edits that you do now may be incompatible with the released version.
So, by all means play with it and test it but don’t use it for edits that you care about.
I got it. Thanks!
Hello. Sorry for the really basic question, but could you please point me to a (surely already existing) guide that will tell me how do I install your build to test without affecting my main install? I use Windows and am afraid that I will end up with two programs simply called ‘darktable’ and not know which is which, or that this build overwrites my current (5.6) installation. I wish to create a brand new database (image library) only for the test purpose that I can later uninstall cleanly, so that I am back to my usual working 5.6 version.
Thank you for this work, looking forward to try it out so I can send some feedback.
See here.
Prebuilt is here… Release Darktable nightly build 20260810 · masterpiga/darktable · GitHub
You can also find how to build it yourself…look in the packaging directory and then WIndows on the official repo for darktable. The instructions are simple and once you set it up its basically running a script each time for a new or updated build.
To build this version you would just need to clone and checkout the branch that @Masterpiga is using for this extra code…
The link shared by @mino should cover the main considerations…ie set it to run in a specified config folder which you can delete once you are done testings…this will contain all the config files and info for your test version and turn off xmp writing so that you don’t accidentally corrupt your main installs edits…I have 5 versions running right now with various features…if I was more clever with git I could likely have one custom one but doing that might create errors from that process that are unique and not helpful to test and integrate the specific ones found by keeping each separate and unique…
Thank you, @Masterpiga, for all your efforts to improve on dt.
This work on mask etc. is tremendous. I really hope it will lead to a change in a new dt version.
Will it be more immediately understandable (for a beginner) and not need an explanation (as @don51’s request seems to illustrate), if “bypass” is rather called “disabled”?
I think so.
Hello @Masterpiga. I had a chance to play with the flexi panel, and have some notes - in no particular order. The notes are based on this playraw edit.
Phew, sorry for this mess of a brain dump. I really hope this might be a little helpful. I am keen to do more testing if you think it’s worth it?
Thanks for the great feedback, @lkzk!
Yes, you are right. I did not really think too much about the shortcuts. In this version I just wanted to expose and test the functionality. As the main features are settled, I will start iterating on the details.
(With “details” I do not mean that it is not important, but rather things that can be changed quickly and iteratively.)
You can do that with SHIFT+Click already. There was an icon for that before. I removed it because the UI was too cluttered. The plan is to expose all the actions available for each row in the row icon menu (i.e., the one that you use to change the between-groups operator).
It still works (at least, it should). It also works if you press c when you are over the button to add a parametric channel, as well as if you do on the slider after adding it to the mask. You may have to wait a fraction of a sec before the overlay shows up.
Noted.
I would use a raster mask for that. The mask will be always be whatever the other module publishes, which changes with every edit.
Yes, that is to to be expected. It is a huge code change and at this point I am not aiming for perfection, just for something that works well enough to test the idea.
By all means, this is all super useful, thank you so much! 
I will do more testing and try to come up with more specific useful notes.
In the meantime:
Amazing, this is super useful. Maybe we can have a little chevron to show this additional detail panel next to the solo mode icon - this could allow to expand and collapse the extra sliders? I really think it would help to surface this functionality.
I might be doing something wrong, but when I mouse over the hz “rainbow” and hit ‘c’ I see this:
…while I would expect this (accessing the same via the classic parametric mask):
Thanks again. I’ll think more carefully through the specific questions you asked in your initial post.
EDIT: couple more things I am noticing:
It is a soft union, useful to prevent feathered regions to create harsh transitions. It is mentioned in the README and in the user docs.
Indeed, it is a no op. As explained in the docs, each group is a bag-of-elements, the order does not matter. Maybe I should disable within-group reordering altogether, but then some people may like to rearrange the elements in a way that makes it easier for their mental model, so I would keep it around.
Yes, there used to be handle to reorder. Removed (like the chevron for the properties) to reduce clutter. As I commented in a post above, darktable’s philosophy is to try and avoid cramming everything possible functionality in a UI widget. Some things, you are just supposed to know them.
This is not necessarily a philosophy that I agree with (as a matter of fact, I mostly disagree), but in the specific instance I kind of do agree. Mask editing is very click and action intensive. If you want to expose every functionality through the UI, you end up with something that is either (1) very nosy and crowded, or (2) very click intensive or with long menus, where you have to click five times or navigate through list of options to get anything done. The alternative would be to reduce functionalities, but this would definitely be not darktable’s way 
So, there needs to be some judgment call being made as to which functionality to expose in the UI and how.
I do not claim that the choices that I made are the best ones, and they are certainly far from final, but the end result will necessarily include some compromises wrt what has a UI widget and what does not, because the main goal is to have something that is quick and efficient to use for those who know how to use it, while still being reasonably self-documenting and discoverable.
So, feedback in the form of
is not very useful, in the sense that we all agree that, ideally, all functions would be exposed via an explicit UI widget. OTOH, feedback like
is very useful because it gives me concrete elements to weigh when I have to decide how to allocate to UI real estate.
In the previous update, I dropped the “properties” chevron icon and the separate icon for parametric mask rows. This freed some space on each element’s row header, which allowed me to move the parametric slider directly in the row header, which saves a lot of vertical space (the 2nd screenshot is wider, but this is just by accident):
Before:
After:
So, we pay a discoverability penalty for a real estate advantage. As everybody here is very passionate about vertical real estate, I think that this is a net win.
A counterpoint thought on that particular example…
I’m passionate about image real estate: I like it to be as big as possible. This means that I hide, or keep to a minimum, all other areas. The downside to this is precision in slider/etc movement over a shorter distance. This might particularly apply to a parametric slider: things get very close.
For the long-term details list…
About the ‘c’ visualisation shortcut: it also works with CTRL-C. This is a nuisance: one cannot copy the history if a parametric mask instance is open.
Thought of reporting it as a bug. But with bigger changes sometime in the future (maybe), I hadn’t bothered
I was wondering, if these parametric mask sliders of Hue, Chroma, etc. could also get this right-click fine-adjustment mode, as we have if for exposure and many others.
I often have pictures where I need to adjust some of these four pickers by a very tiny amount, which is not so easy. In the current parametric mask controls, there are also the resulting numbers output in the GUI, but one is not able to edit them (which would also be fine from my point of view).
How do others think of this? Or am I alone with this “need”?
Good idea! It’s something I’ve only recently started to use, and I suspect it isn’t well known. I hadn’t thought of it for parametric masking. No of course: it isn’t there.
Whether it’s the answer or not I don’t know. But tiny adjustments in this instance are critical, and very hard to make
You might know this already, but right click does ‘select’ or deselect the individual triangle/value so you can use mouse wheel to increase/decrease it without moving it first.
Is it more suitable than a precise input box? Not sure.
Thanks, that is useful to know. But I have wanted the value adjustment scrubber popup so I could see which side of the mask was responsible for a mask error without permanently changing the value, or changing it but then being able to make the change more subtle before the value is committed. For this use case, one click to make an aggressive change works (as it will be undone), and the value scrubbing popup is even better, but mouse wheel adjustment doesn’t fit the workflow.
One issue: if right click brings up the value scrubbing popup, it would break the triangle select functionality you just mentioned. Does the scrubber also work with the mouse wheel?
Oh I didn’t know that. I realized that the triangle gets highlighted when right-clicking it, but I didn’t know that mouse wheel would change its value. Good to know and thanks for pointing to that. I typically work with the trackpad of my laptop, middle-mouse and pointing-stick (which should be emulating the wheel) scrolls the list of modules 
Unfortunately no: you move the value by dragging the indicator with the mouse. The mouse wheel just changes the “zoom” level of the scrubber pad, i.e. the current min-max span. So perhaps, it could be worth to introduce key-accelerators here, to also allow for adjusting with the mouse wheel.
Mouse wheel does nothing for me here. I think it is because I have it set to scroll modules: I do that all the time, so this is totally set in my use pattern
Is this scrubber pad are we talking about the “whisker control popup”…if so at least with the mouse it has two nice features…ie the proportionality ie the further out you are on the whisker the more precise the adjustment and secondly if you enable it and just move the mouse pointer without left clicking it makes the changes visually so you can move it around until you like it but if you don’t left click and just hit esc the value is not changed, if you do left click when moving the whisker then the change will move the value when you let go…
Edit acutually even better you don’t need to hit esc just moving the mouse pointer outside the popup closes it with no change applied…
Edit
Regular sliders have a modifier as well for the mouse wheel… Shift does 10x the default single increment and ctrl does 0.1 for finer increments… Finally preferences has a setting to scale the slider to the min and max… I am not exactly sure but likely it just adjusts the default based on the span of the slider if you go beyond the soft limits??
Ikeep meaning to check the documentation to see what the “official” name for it is, but I don’t think it can be any better than Whisker Control.
I think it could be a great help in the parametric panels. But the response of the image preview or (possibly more important) the yellow show-mask overlay needs to be very reponsive.
