Ya I can see that I was really at this point only considering the color science setting so that this could then let the module work as if it was free to set defaults and not lock in some values created during the preset… maybe that is what you mean as well…
I see the same values ie 4 for RW and -8 for RB for every image and if I copy the image and tweak exposure to some value no change… so if I had a preset with those values and color science v5 I am not sure what the issue would be but I think I am missing something… obviously if I set the preset with different b and w points those would be applied and might be worse than the default filmic starting points…
But filmic doesn’t seem to do anything for me based on inital exposure, should it be ??
Maybe hardness ends up at the wrong value applying filmic this way??
filmic rgb sets up some defaults based on the exif exposure compensation value (the same value that’s used to “compensate camera exposure” in the exposure module). See here.
1 Like
Thanks…I will have to pay more attension…I don’t ever seem to see any change in starting values…I can see the math so it should impact B and W rel starting points …
@kofa Ignore me I’m a twit…its because I don’t auto apply filmic. I have the settings in workflow set to none and modern… I use filmic when I need to but not out of the gate I like to see the image just white balanced for the most part before I start… that’s why I don’t see the small recalculations…I was just trying to wipe the history and apply filmic to get it to do it and it doesn’t … but of course it has now occurred to me why…
@christoph I was doing some experiments comparing v5 with v6 and found an image where I also prefer the results from v5 better than from v6. See Comparing filmic color science v5/v6
Hi Andrew, I stand corrected. I tried creating a preset for filmic V5 on my windows computer at home and it worked. Still, I have no problem using V6 and will continue to do so.
In the original post it stated > Creating a new preset and apply it automatically to each photo doesn’t do the trick as filmic rgb doesn’t use the same values for white relative exposure and black relative exposure for each photo. Using a preset would also overwrite these values.< I have not tested this yet.
It’s only triggered if you have the workflow set to scene referred on the initial opening or if you reset the image but resetting will use V6 I believe…I wasn’t seeing it as I have the workflow set to none. I like to see the image without filmic. In that case when I would apply it the values for white and black were always the same for each image…
After comparing v5 vs v6 in a lot of images, I also prefer version 5 way a lot more than version 6.
Does someone know why v6 was developed? What should be theoretically better over v5?
It was developed in part as many people always complained about filmic being dull and too desaturated…
I seem to be missing this panel - please explain where/what it is. Google is not much help; it tells me: "Xilica’s XTouchApp enables users to control their device(s) using any Mac, Pac, iOS or Android device. ", none of which devices I have that also run dt.
I’ll join your opinion. Filmic RGB has not appropriate default settings and that’s why darktable produces bad results out of the box.
Here is an explanation of my thoughts. The default result should be either realistic or similar to in-camera jpg.
The realistic look can be easily convinced: just take a picture of some “permanent” close to your computer scene, process it and compare the look of the scene with the picture in darktable on a calibrated monitor with your eyes. The scene can be LDR to suit the most simple and common case.
This comparison is enough accurate as it uses the comparison (difference) method, which can exclude personal preferences on the look of a scene.
And the results from darktable with default settings have been non-realistic for some time. They look dull. They were dull with V4, V5 and V6. I had to make my preset to obtain realistic looking results.
Concerning V6. The most close-to-reality results with standard color matrix are:
Contrast 1.3
Latitude 50%
Extreme luminance saturation +100%
Target black luminance 0
Preserve chrominance: Luminance Y or RGB euclidean norm.
Both contrasts: safe
Exposure + 1.15
You know that you’re not forced to use filmic, right? You can even use base curve in place of that module (near the end of the pipeline). That may give results closer to the in-camera look.
1 Like
OK, so you used exactly 3 modules beyond the required basics:
- denoise
- exposure
- filmic
If that is indeed how you want to use darktable, the best solution might be indeed to switch to using the basecurve.
I use at least two other modules as a matter of course :
- color balance RGB to add saturation over the whole image
- diffuse and sharpen: demosaic sharpening
From there, often a seond copy of diffuse & sharpen for local contrast ,
tone equaliser, extra copies of color calibration, … as needed.
Your insistence on doing everything with filmic means that in your sample image, you completely crush the blacks (and whites, but there aren’t too many of those). Try this:
- color calibration : daylight 5500K
- filmic: contrast 1.00, black reference -5.9EV, white ref. +4.24EV
- diffuse & sharpen: “add local contrast” preset, increase iterations to 20 (or so; or use “local contrast module”)
- colour balance with “basic colorfulness: standard” preset.
And keep in mind that the scene was dull! very flat, diffuse light, so no shadows to give texture and depth.
Or perhaps I’m just wasting my time replying to a troll (
)
1 Like
Of course, I’m not . But it is obvious a program should produce good results out of the box. They can be either realistic or in-camera like, but not the ones they are now. That is not mine opinion, I share it with some other people who use darktable.
P.S. I’be been using darktable with my handmade color matrices and base curves, then Filmic and own style for a long time (from the beginning of the program). But it would be better to have appropriate results for a regular user.
I find filmic’s default look appropriate. Now, with the auto-tuner working better, it’s quite easy to adjust. Since v6, I find I often prefer the look of Luminance Y or no chroma preservation. Not as contrary as most the in-camera look, for sure.
Thanks for the hint with lower black reference. I’m now playing again with 200% increase of saturation at the edges.
P.S. I don’t think it is correct to enable 20% saturation increase in the other module (color balance RGB) all the time to get realistic results.