fimic rgb saturation curve

i see in v3 to v5 this curve is present, in v6 it is not, as you said.
Thanks

If you are curious

Yes, I am certainly curious, but this much expertise to do anything with this text overwhelms me.

My impression grows that dt is very difficult to understand. But I hope that you can still use it without being a color scientist or computer scientist.

I’m neither a colour scientist or a computer scientist. I find it easy to use (but I do watch lots of video tutorials and RTFM every new release).

3 Likes

Maybe you will find the new sigmoid module easier.
It’s also possible to use darktable in display-referred mode, with a base curve. You still have the white balance, sharpening, lens correction, and so on. You are not forced to use it in one particular way, using one set of modules.
And then there are many other tools. I often like the simplicity of Filmulator, and many praise ART. And there are others: PIXLS.US - Software

1 Like

filmic is quite simple, just ignore most of the sliders.

  1. use exposure to set the overall brightness, concentrating on midtones;
  2. set white and black points in filmic according to taste
  3. set the contrast to taste
    With higher contrast, or unbalanced white/black relative exposure, the curve may clip. In that case, switch to the advanced tab, and set the contrast in shadows/highlights setting(s) to safe.
  4. If you are not pleased with the colours, scroll through the preserve chrominance options. It only takes a few seconds.
  5. If still unhappy, try a different color science (one of v5 and v6 will most probably work well).

If this is till too complicated, try sigmoid. It has a contrast slider, one to shift contrast towards darker or lighter tones (skew), one color processing selector, and if you choose per channel as the color processing option, another slider that you drag to see if some setting gives you more pleasing colours.

If still unhappy, try base curve, possibly with a fusion setting. You can find a well set-up base curve in the XMP attached here: filmic v6 loss of contrast - #19 by Leniwiec. @Leniwiec swears by the base curve.

Or try some styles – but not from dtstyles.net, which is outdated.
A selection of Darktable Styles – One Camera One Lens
Editing Made Simple In Darktable. A Video. – One Camera One Lens

10 Likes

You can, but is it worth the effort? Also the effects of using current implementation of filmic rgb can be very underwhelming in my opinion.
I personally ditched filmic and use one of display workflow based curves - makes much more sense from user convenience perspective.

I know many feel that way, and I don’t doubt they have good reason for that. I just find it interesting. For me, the scene-referred workflow really made things easier. Previously, I always had to be careful to avoid clipping. Of course, then we didn’t have tone equalizer, either. Glad it works for you (and for many others).

The important point is that you can get results that you like. :+1:

2 Likes

Great way to shape it. Tone curves are just curves. Pick one, or two… :laughing:

I’ll add one to your list: control-point curve. There’s only so much you can control with the parameters feeding some sort of curve equation; the control point curve lets you shape the tone mapping anyway you want, even with goofiness like negative slope. I actually have done this a couple of times, pulling the curve back down in the mid-range so it can influence something like a bright sky, knowing that where I pulled it negative had very few pixels…

For high DR images I’ll sometimes use two curves: 1) a log-sort of curve to lift everything out of the dark well, and 2) a subsequent control point curve to shape the mapping to my liking. Makes it easier to deal with the darker tones. This I do in rawproc; it may now be do-able in dt with the recent changes to allow manipulation of the pipeline, although dt may not have a UI-usable log curve…

1 Like

Would you say that setting data to 0-1 and then shaping that is the same as a curve anchored at some point in the raw data that gets labelled middle gray and then mapping around that…

I guess what I am always trying to understand is what data is being used … a curve is a curve is a curve but is it a “normalized” curve or unbounded???

One option would be to lower exposure, and apply the tone curve tool with its log preset:
image

Then apply another tone curve on top to add contrast etc.

Another is to try filmic with something like the attached preset: no contrast, blend in lightness-only mode. Auto-pick the black/white point, then add a tone curve for contrast.
filmicrgb_filmic-log.dtpreset (1.0 KB)

2 Likes

Thank you all for your thoughts. I probably expressed myself in a misleading way. I am doing quite well with fimic rgb. I just meant reading the recommended stuff (Filmic v6, etc.) overwhelms me tremendously. But I’m not questioning the quality of filmic rgb.

2 Likes

Not at all. But I think the middle-gray thing is a parametric artifact, a way AP thought would make filmic more understandable.

I struggle a bit with the “unbounded” thing. After all, the sensor saturation wall is really a very hard upper bound, which has to be scaled up to make the data fit the 16-bit integer range. 0.0-1.0 floating point is a little goofier, in rawproc I start with the raw data scaled to an equivalent magnitude in the 0.0-1.0 range, so the starting histogram looks like it would if the data were still in a 16-bit container. But that’s beside the point…

In rawproc I did have to modify my control-point curve to not truncate at 1.0. So, when I drag the upper control point down the right side of the window, there’s an implicit continuation of the spine curve past 1.0 that is used to map data. Making it available for dragging back into black-white range with a subsequent exposure or normalization tool. But I don’t do that sort of thing in practice…

Feeling like this is a sort of stream-of-consciousness; hope it helps provide context…

2 Likes

Oh, indeed I swear :smiley: … providing it is used with v.3 pixelpipe order :wink:

And it doesn’t mean I’m complaining about Filmic - I perfectly understand what benefts it brings, why it is needed, what it does, why it is newer and future-proof and so on.

Moreover, I tinker with it from the day one on regular basis, I like (!) and prefer to use it with my black & white female portraits (subtle, gentle and overall nice tonality, easier to hide skin defects, wrinkles in harsh light), but…

It’s just me and my preferences - somehow for general, colour photos I prefer the pop and the colours of the Base curve and I control clipping with Tone EQ before the image is tonemapped by the Base curve.

Another convenient for me thing is, that Base curve doesn’t care about where the black point starts or white point ends and so on. I don’t have to anchor the exposure around middle grey, which all leads to simple fact, that working with it resembles working with my mirrorless camera - if I see an image on the camera’s display and the exposure looks good to me, I know that after applying my base curve it will look the same – again – to me :smiley:

So:

  1. I respect the decisions made by developers.
  2. I understand why they made it.
  3. I use new provided tools.
  4. But I also follow my taste and I’ve chosen my beloved tools & workflow.

All of this may be also usefull for @strumpfli and Filmic RGB as a default module - Software / darktable - discuss.pixls.us

Thanks for mentioning my preset :wink:

2 Likes

Is this the new or old position… I am assuming the later one in the pipeline…but maybe I am wrong…

The new one - for scene referred.
The Base curve is at the top, just like Filmic or Sigmoid :relaxed:

I just choose “none” as pixel workflow default, have a look at the image and then decide on the set of tools desired - presets do the job!

Same… although I see the base curve has those color preservation modes as well… I have not really tried them to see if they have as much impact as they do with filmic…

Unfortunately, they don’t fit my taste whatsoever :woozy_face:

Is there a PR or discussion about these new defaults somewhere? I found some PRs on Github about the highlight reconstruction, but I can’t find anything about the reasoning behind the new contrast and latitude defaults.

As far as I recall it was a decision made by Aurelien and adopted I can’t remember where the initial discussion if any was. Basically I use the autopicker now and leverage the exposure module. I do a full image at 50% and this works more times than not and then if not I select a region that should be 50% and select that and work from there… As for filmic I start without it and decide if it is needed and then if so I prefer v5 and how it works so I would use that if using filmic…the gamut handcuffs in v6 might be great but I don’t like them.

Edit I do recall it happened right when AP forked DT and created R/darktable, now called Ansel…

1 Like