Tentative RGB workflow for DT3

Hello

I’ve been lurking here for a while trying to wrap my head around the new RGB workflow, and after reading a bunch of posts here I’ve come up with this checklist which seems to mostly work. Not all steps are necessary for all photos.

It is very much a learning process, and I’d be very happy to hear what the rest of you do.

  1. exposure correction before filmic
    • center and correct histogram roughly focusing on mid-tones
  2. white balance if necessary
  3. crop and rotate to correct composition, if necessary
  4. filmic rgb RAW correction, dynamic range, and contrast
    • set white exposure or auto-set with the colour picker on the brightest area
    • set black exposure or auto-set with the colour picker on the darkest area
    • adjust contrast on the scene tab to your liking
    • reduce the latitude to avoid the curve clipping
    • lower the grey middle point reference to brighten the picture globally if necessary
  5. channel mixer or color balance for BW conversion
    • in channel mixer use film presets
    • in color balance desaturate
  6. color balance for
    • adjusting saturation and contrast
    • color corrections (instead of velvia/vibrance)
    • black, white and grey point (instead of levels)
    • separate instances with parametric masks (instead of color zones)
  7. local contrast
    • it is possible to set shadow and highlight to zero
  8. contrast equalizer for clarity and saturation
    • increase luma at all frequencies for clarity (there’s a preset)
    • increase chroma at all frequencies to saturate
    • limit to midtones, using a parametric input mask
  9. tone equalizer instead of shadows and highlights and tone curve
    • on the masking tab, below use the sliders to adjust the histogram on the advanced tab
    • use the mouse wheel on the image to dodge and burn
  10. denoise (profiled) if necessary
8 Likes

Hallojsa René & welcome!

Your workflow sounds very interesting!
Thank you.

Have fun!
Claes in Lund, Sweden

2 Likes

In Filmic RGB, if you’re going to adjust the gray point, do it first, not last. This way your tones are locked in before you move on.

1 Like

I set exposure as you said. In filmic, I set middle grey to 18% and the rest of filmic steps the same.

If I need brighten it, I modify exposure again.

Why 18%? Because of this :

Regards

2 Likes

I’ve read those posts, and experimented a bit, but I found it harder that way. I’ve studied enough math in my past to understand that it is right, but I’ve failed to get as good results that way.

Lots to learn :yum:

I’d add also lens correction.

I still can’t wrap my head around parametric masks with color balance for “better” color zones (hell, i still haven’t figured out color zones). And how do you use contrast equalizer?

with denoise profiled - one has to adjust strength if increasing the exposure.

3 Likes

One little thought…what you crop may have different characteristics than what you keep after cropping. I like to do any rotation and perspective correction first, then crop, and then get down to business.

4 Likes

Thanks a lot for your input. Truth be told, your workflow is exactly same as mine which I have been able to gather together watching tons of youtube clips, reading fora…

1 Like

That’s exactly the problem. You need to read and watch loads, take notes and be systematic, just to arrive at a basic starting point.

That’s not a problem. That’s certainly how I learned development back in the day, only back then there were no videos, and precious few books! Today you can get the source code for darktable, and read the docs and watch videos, and if that’s not clear you can post a question here, or reddit, or the mailing list (too many places! maybe just stick with this forum!) and the developers/experienced users help. As long as you have time, it’s all good!

Regarding your question, one of the developers wrote this:

You don’t have to worry about the order:

@seindal, this is helpful. I’m still learning filmic but I’ve made some comments in normal type against yours in italic.

1. exposure correction before filmic

center and correct histogram roughly focusing on mid-tones

@anon41087856 , in the dt 3.0 manual on workflow, step 1 is adjust exposure then step 2 gets into filmic. When adjusting exposure, should filmic (presumably default settings) be on or off? And if On, I think you recommend 9% rather than 18% for outdoor daylight shots, yes?

2. white balance if necessary

It should be correct before adjusting filmic.

3. crop and rotate to correct composition, if necessary

4. filmic rgb RAW correction, dynamic range, and contrast
set white exposure or auto-set with the colour picker on the brightest area
set black exposure or auto-set with the colour picker on the darkest area

At this point, what to do about dynamic range scaling?

@anon41087856 , this section in the manual mainly explains about auto-tune, it would be nice if there was a bit more about using this slider manually. There’s a bit in the “remap any dynamic range” youtube about this (at about 1hr19min).

So after setting black exposure, is this a good time to assess clipping, and adjust DRS accordingly?

adjust contrast on the scene tab to your liking

reduce the latitude to avoid the curve clipping

Adjusting shadow/highlight balance might also help with clipping. Or go back to Scene to adjust DR scaling slider?

lower the grey middle point reference to brighten the picture globally if necessary

1 Like

Setting exposure with filmic off is fine. If it’s on, ensure scene grey = display grey = 18.45%. It’s no big deal if that condition is not ensured, but it’s a cleaner workflow.

Nothing. Setting scene black and white points is the dynamic range scaling. From these values, filmic remaps white to 100% display luminance and black to 0% display luminance. All the rest is just adding contrast and taking care of oversaturations to you taste.

The DR is already set since it’s white - black.

Curve clipping is a look thing, it should be unclipped in the look parameters. Please stop messing around with scene parameters as if they were some artistic parameters, they are not.

  • Scene black is the lowest light emission on your scene, as captured by your camera.
  • Scene white is the highest light emission on your scene, as captured by your camera.
  • The difference between both is the scene dynamic range.
  • Your camera will capture them both if they fall into its sensor dynamic range. But usually, the scene dynamic range is larger than the camera one, so one or both scene bounds will be clipped to the camera dynamic range bounds.

The “scene” tab is meant to be set first with absolute values depending only on the light emissions, as captured by the camera, for the current picture. Then, the “look” tab is meant to do whatever you want, independently from the scene. The point is, a log tone mapping (scene to display with no look) looks very dull and not very plausible to the eye, so the look is a sweetener to allow a bit more contrast in the middle of the dynamic range (= the latitude).

Whatever mistake is introduced at the look level has to be fixed at the look level. Wanting to add more contrast than the algo permits doesn’t change your scene light emissions.

If anything, filmic showed that people struggle with the basic concepts of white and black, which means they don’t get the very concept of exposure either, which is quite concerning because it’s really all what the camera is about.

So, “black” and “white” are bounds of the particular (output or input) medium you are working with. They are purely symbolic, because their absolute luminance depends on each particular medium, and “black” or “white” doesn’t tell how many candela/m², which we need to know at some point. “Black” is still some small amount of light, you call it black because you don’t see much and don’t care much about light emissions darker than this value.

Usually, digital image processing doesn’t care about absolute luminances and lets that to the color management system. So we are used to black = 0, white = 1. But that’s just an encoding, scaled and offset for our own convenience. It doesn’t mean anything in itself.

Filmic treats black as what it is : not a zero, but a very small, non-zero, light energy which lies at some distance from another light energy, called white because it will be later be mapped to the output medium white, but is actually just the saturation threshold of the sensor (so, the maximum of some recording range).

What your camera records is some values proportional to some scale (it’s dynamic range). In this scale, there is something more than the bounds : the middle grey. That name is confusing, because the camera doesn’t actually anchor the middle grey in the middle of the range. What’s even more confusing, is if you wanted to paint a middle-grey between some white and some black, you would have to mix 20% of black paint with 80% of white paint.

Why middle grey is so important is because it lies straight in the middle of our perceptual range. That’s the middle we are talking about. The problem is, when you open a camera raw file, it can be anywhere, and the user is the only one that can decide what is to be considered grey, because it’s a perceptual middle between some more or less dark black and more or less bright white.

Correcting middle grey without messing the whole (linear) encoding is done by an exposure compensation, which rescales the whole dynamic range by some factor, so that middle grey falls at 18%, where the display expects it. The problem is, the display also expects diffuse white at 100%, and your scaling has probably pushed it between 200% and 800%. Here goes your display-referred pipeline, built in the 1980’s around the idea that anything brighter than diffuse white has no right to exist in the picture.

What filmic does is anchoring the scene grey to the display grey, and bending the bounds of the dynamic range around that value, so the 18% middle grey passes unaffected to the display, but the white and black bounds are force-fitted into 0% and 100%. But, while the “black” and “white” meaning for the display are pretty well defined (white = paper with no ink, or LED at full power, black = full ink on paper, or LED shutdown), for your camera, they can be anything. Hence why you have to tell filmic “this much lightness from the scene is black” and “this much lightness from the scene is white”.

Not sure if I’m clarifying things or just increase the confusion, though.

9 Likes

Maybe copy and paste that whole post as the text of a tooltip which appears as you hover over Grey Middle Point!

Then have the Holy Church of the Intuitive UX on my back ? No thanks :smiley:

4 Likes

Reading is the least intuitive activity.

Further clarified for me, thanks!

Nah, Into documentation and my bookmarks it goes!

Those are blasphemers! Nothing is intuitive untill it’s learned! Intuitive UX is the one which streamlines 80% of tasks into minimal work amount while allowing remaining 20% possible without hindrances!

Judging by the amount of “phtographers” who buy newest CameraCompany CameraBody-Mark0 and then ask on reddit/facebook etc “How to do X” where X= a thing that’s explained in the freaking manual coming with CameraBody-Mark0 in “Getting Started” section… I tend to agree with you.
And I have to ask: WHY people don’t read manuals? It doesn’t hurt…

Every time you start darktable, this should appear as Star Wars scrolling text with music by John Williams :upside_down_face:

6 Likes

@anon41087856, Aurelien, thanks for this. I’m sorry you’ve had to write such a lot!

I get the message there’s no need to adjust the D.R. scaling slider (although that is what you did in that section of video I mentioned!, in connection with local contrast and clipping. You see, I’m following your teaching, and now you tell me off! ). It’s in the GUI, what’s it for?

Mid grey is clearly important but I’m struggling with it! I have the image before me, and I need to envisage the actual scene when I took the photo, perhaps months ago, and judge what part looked half way between black and white. Then I can sample it on screen. Tricky. So inevitably I go iterative and can’t resist coming back to tweak mid grey or exposure in case that makes it look better (to me).

But thanks again. I really like what filmic does.

to easily move both white and black sliders at the same time, while keeping the ratio consistent.

set it to 18.45 on both tabs and forget about it :wink:

It’ll take practice. If you’re shooting same type of photos in similar lightning conditions it’ll be easier. If not and like me you’re shooting “anything” then it’ll take time and practice. I find that for conditions I worked previously, my workflow with filmic is significantly faster with more predictable results than any other method (except when filmic fails on me, but then I either don’t try and do other things or leave it for some time in the future)

The dynamic range scaling factor is only a way to quickely adjust scene white and black in a symmetrical way and at the same time, as a shortcut. It is mostly useful when you use the auto-tuning and don’t have true white to sample in your image, so you give some safety margin to the tuning (which tries to match max(RGB) in the picture to pure white). Of course you can decide to clip white or black past a certain threshold if you decide there is no important detail in there and that preserving the local contrast is more important.

So force yourself to set the exposure with filmic disabled, set darktable to the grey theme (which is middle grey display-referred), and adjust the image exposure so the average brightness matches approximately the grey UI. Then setup the white and black mapping in filmic.

Middle grey setting is just a mean to an end : it’s intended to provide a reference for a good, legible, average exposure in the picture. Don’t put too much meaning in it, and don’t try to edit by the numbers. If it looks good, it’s good, middle grey or not.

Then, as @johnny-bit said, let grey = 18% on scene and display, and don’t bother.

1 Like