For global tonemapping there’s no way I’d place curves before/under filmic rgb, but for the local edits I’d like to do it.
Ya that’s likely the criteria…and if the assumption is that filmic is used for global tonemapping then the default order makes sense for local edits…
Hard to say (you would have to ask them), but if clipping can practically occur (even to a small extent) then it is unlikely to be a linear space implemented with floating- or fixed point arithmetic (for reasonable choices of the latter).
The biggest hurdle for me was realizing that I can have multiple instances of the same module, each doing a tiny bit, maybe masked. Then I can combine the same 10-15 module types to build something compex. I think this is atypical for software like this, and requires a good intuitive understanding and lots of practice. I am learning a lot from the Play Raw edits though.
If darktable had been organized this way when I started all this, I might not have written rawproc. Discretion with the order of operations has been one of the most poignant learning situations I’ve had in learning digital imaging.
1 Like
shadows & highlights is not (yet) deprecated
Oops, I misread something… That will teach me to check all possibilities (I thought it had been completely removed for new edits after deprecation)
When I am lazy and just tweaking some phone snaps I will often use it …bilateral so no halos and alsmost always with an adjustment of the white point slider…surprisingly you can get a nice quick fix with it …if not then I ditch it and go to the tone eq.
Me too (shh don’t tell anyone), even on raw edits. I use tone curve too but the key thing is that because I get 95% of the way there in the scene-referred part of the pipeline I don’t need to push the display-referred modules nearly so much, and they behave much better.
Exactly… I use 3 or 4 “old” modules sparingly but for small tweaks… tools in the tool box…
1 Like
That’s bullshit. Large color space doesn’t imply anything on whether the data are bounded (display-referred) or unbounded (scene-referred). That’s word play to say that the chromaticity space is large. But chromaticity is orthogonal to luminance. And if luminance is bounded, back to square one : it’s display-referred, no matter how large the gamut.
5 Likes
Capture One has five sliders that can reach “beyond white” and “below black”: Exposure, White, Highlights, Shadows, and Black. Everything else can not pull colored highlights from white, nor detail from black.
As a user of both Darktable and Capture One, this puts a hard constraint on my editing. In C1, if your dynamic range does not fit into the display space, you must use the above spacial (!) adjustments to make it fit. No curve, no contrast, no levels can help you. And worse, all of these sliders can produce halos.
But I don’t want to overstate it. It’s not hard to live with a system like that. In some ways it’s easier, actually. Less choice. Only one tool for the job.
I jump back and forth between easy Capture One and powerful Darktable. Sometimes I get hung up on an image in one editor, then I switch editor, and everything becomes clear. Then I switch back and understand what to do.
One thing though, that Capture One does really well: they have very very good video tutorials. Short, tightly edited, to the point. Of course I like Aurelien’s rambling, nerdy streams a whole lot, but having full-time staff just doing tutorials must be great.
1 Like
This is what happens when you pay people, they have the means and time to go for higher production value.
2 Likes
As long as I read about Capture one workflow (really not much info about it) C1 processes in the camera color space.
I uses the profile of the camera and transform the data to generate color using that profile.
Then it works there, with the bounds i gamut that has the camera and only converts to the output profile at the end.
There are three approaches (or strategies) to processing:
- Work in the capture device space (as it seems C1 does) and convert at the end. It seems to be a good strategy to keep all the colors you captured, but you can generate a lot of colors that you are not viewing in your monitor and final output device.
- Convert from the camera to an intermediate space (be it linear and unbounded or not linear) wide enough to keep most of the colors capured by devices and to keep colors for monitors and printing, and convert from there to the ouput device profile.
LR and ACR use a variant of ProPhoto, which is wider than REC 2020 used in Darktable (not sure that is good for a integer processing pipeling, but I suppose 16 bits are more than enough to not produce banding in ProPhoto). The gamut is the same as ProPhoto but they do not use the same gamma not sure if it is linear (1.0) but more linear than the sRGB. - Convert from camera to output device color space and work there. This aproach lets you controlled with detail the edition and do not generate colors out of range (you still have the problem of visualizing them if your display device for edition is different in gamut than the final output device). The best result and control of colors for the output device, but you have to reedit when changing final device.
Most program seem to use an intermediate color space for edition in order to have a reference standard to process images.
1 Like
Are you sure this is the case?
Consider a simple channel mixer. Suppose you have a pixel:
RGB = 100; 10; 20
And your green channel’s RGB mixing is set to R=-0.15; G = 1; B = 0
Your output green will be -0.15 * 100 + 1 * 10 + 0 = -5.
What RGB space you use does not really matter here. Or am I mistaken?
I did not mean that the conversion could not generate out of gamut colors that must be converted somehow, I meant during the edition, as you are working already in the destination space and there is no final conversion.
Using an intermediate space may disregard colors present in the capture (or that you want to generate) and in the output device but not in your color space.
With a ProPhoto space it is not probable, as it is quite wide and probably wider than current output devices (wider in some zones than human vision, narrower in other parts).
I am not advocating the use of this approach, it was the workflow at the beginning of image processing, when programs did not make color corrections, you were working in your display color space, indeed. Nowdays, I think it is better to not work like that, I don’t know if somebody does for special cases or devices.
You are talking there about linear conversions, using a matrix transform from camera to RGB.
But color profiles are not linear, and C1 is supposed to use a color profile of the camera.
You can provide one or it will use a canned profile (depending on the style you choose).
No. I meant a channel mixer, which is a well known, widely used editing tool (backed by a matrix multiplication, for sure).
https://docs.gimp.org/2.10/en/gimp-filter-channel-mixer.html
1 Like
Yes of course you can use tools that generate OOG colors, as you can create them in Darktable, but they will be inmediatly clipped and you will have a feedbackup of that and see inmediatly the effect that is generating the problem, not later in the workflow when you convert it to the final output profile.
If the output is the monitor you are working on you have a visual feedback, if not, you need to activate alerts.
But not a good way of working, I think, usually, as you have to edit it exprofesso for that device.
Using the camera or capture device space is apealing in order to keep the maximum of the captured colors (today they won’t be shown in monitors or printers, but who knows in the future).
Capture One seem to do that, and gets good results, I like it more than the LR.
But not sure if it is the best way of working. It seems that a standarized color space to work in is more contrlable.
Working profiles are linear. The TRC is distinct from the primaries…
Color profile transformation from a capture device data to the color space, as the filters in camera are quite different from the response of our cells, even when they treat to emulate it.
I don’t think capture one is doing any voodoo colorspace magic…I think they have good ICC files with custom curves …some people love them and some people don’t…
I think they use Lab for PCS??
The nuts and bolts of making one are pretty straightforward…not much different than DT…
https://torger.se/anders/photography/camera-profiling.html#the_easy_way_c1