Trying to figure out RGB (filmic) workflow

Hi everyone,

I have watched a bunch of videos and read a bunch of stuff, but I still can’t seem to get decent results with filmic. I’m don’t have much experience with Darktable but I have watched about 14 hours of videos and played with this file for a couple of weeks with little success.

Using Filmic I can’t seem to pull the detail out of the snow while brightening up the person in this photo. I was actually able to do a decent job using the pre-rgb workflow but I’d like to figure out filmic. Tone equalizer has been much help, but I’ve had some success with local contrast and filmic combined if I mask out the snow in filmic and mask out the opposite in local contrast.

I’ve also tried to use out 3d and a couple of fuji sims for provia but that just seems to annihilate the snow (goes pure white) completely.

Here is a export from lightroom (I’m not trying to say LR is better, this is only to prove that the detail is actually in the Raw file). I believe I did a good job nailing exposure out of camera but can’t seem to get the detail and bringing up my subjects brightness.

I’d really like some advice here because my head is getting sore from running into the same wall over and over again.

Thanks in advance.

(Edit: sorry about not adding a license. Creative Commons Licence
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.)

DSCF8284.RAF (47.8 MB)

1 Like
  1. there are multiple threads for “how to use filmic RGB” already. some are even sticky
  2. if you want to make it a play raw. please assign a license to the photo.
1 Like

I played around a bit with the raw and also could not get the detail in the snow to the level you got in your LR export. I’d be interested in seeing what others can do, but you’ll need to assign a licence. Info on that here: Creative Commons Licence

1 Like

Easy to reproduce using RGB or tone curves. As soon as you turn filmic on all of the highlight detail goes. Preserving highlight detail is the thing I have most trouble with in filmic.

1 Like

Just a note: basic curve was seen as “must have” module, because “you’ve gotta apply curve, right?”. Actually you don’t have to. even filmic rgb isn’t an universal module, so it won’t play nice with some particular images (but is way more configureable).

as @elstoc said - you can try using other modules that can give image it’s curve.

and if you want help with particular image, you have to licence it properly (and if person on the photo has model release that would be good to have too)

Filmic RGB applies a S-like curve. This inevitable means that you will loose some details in highlights and dark shadows. Filmic is not the universal tool for every image. I applied RGB-curve or zone system or local contrast or tone equalizer to your image, no problem to reproduce details in highlights. And as already mentioned by others : please assign a license.

The result you see in the snow is the product of a local tonemapping operator, pushed too far so it starts adding local contrast instead of simply preserving it. You have more details in the snow simply because it’s light grey, not white anymore.

Filmic is a global tonemapping aiming at protecting midtones first, at the expense of highlights, acknowledging the fact human vision is a lot less sensitive to highlights. That’s another strategy, more suited for backlighting. Here you are in plain high-key.

What you should use in dt to reproduce the LR effect is the tone mapping module (using a nasty algo that performs really well in this particular case) in combination with local contrast in local laplacian (which is the underlying algo in LR clarity), and temper filmic defaults.

Here is a rough emulation, with a masked tone mapping to protect shadows (but then the face looks brighter than the snow… I don’t like it):

DSCF8284.RAF.xmp (22,0 Ko)

(I repost unlicensed stuff, I’m such a gangsta)

3 Likes

I’ve reproduced it using filmic rgb, color balance and an extreme use of the contrast equalizer.

@MrKellyGraham Unfortunately I’m not a gangsta, but I promise to share if you provide the license…

1 Like

My legal advisor would chew me out for 1h straight and bill me double :stuck_out_tongue:

Yeah. The point is - tools are different.

I’m not posting anything till the licence’s correct but I’ve had success with parametric mask and contrast equalizer…

As @aurelienpierre said - it’s because snow IS white while the changes done in LR turn it gray. Photographing snow for texture is like photographing white wall with off-white accents: it’s very problematic to keep everything “as the eye sees” since with such bright reflector (snow) our eyes adjust bringing exposure down and using our brain to do in-place tone mapping. You know why people on the snow banks skiing wear sunglasses? It’s not because sun shines so bright, but becase snow reflects so well.

In order to have easier time when processing photos like those, you may want to consider getting polarizing filter - that could optically limit reflected light and bring out a texture (I need to try that, but thanks to climate change there was NO SNOW IN POLAND IN WINTER!!! wfh right?)

2 Likes

Thanks for all the useful replies everyone. I’ve added a license to the file in the original post. Sorry about missing that the first time around.

Sometimes I willing to let snow blow out but in this case I feel it’s weird without any depth information behind the subject.

I’ve being trying to stick to the modules listed as belonging in the new then workflow, to get used to the new way of doing things. But I see know that might have been a mistake.

I look forward to further info now that I’ve included a license.

3 Likes

Photographing a white wall with off-white accents sounds like a reasonable thing to do for plenty of photographers. It’s also quite easy to solve the PP with non filmic workflows.

After my initial experiments with darktable filmic I found that it has built in assumptions, such as the ones were seeing here.

It would be great for the usability of filmic to spell out the assumptions and perhaps give examples of the corner cases, such as this one, where choosing a non filmic approach might be better. Mapping out the corner cases and listing them prominently in docs might help smooth the adoption.

edit: looks like @aurelienpierre did a decent job with filmic

1 Like

If you look at the doc for filmic (the old one), the corner cases and assumptions are already in there. I wrote like 5 pages (DIN A4 size) of stuff that people seems to rediscover every month or so. Same with my videos, where people complain about the length, but I give all the whereabouts. At the end I just feel it’s all lost, I will have to answer that same question again anyway.

2 Likes

There are a few sources of documentation for the old filmic. Please could you link to the one you’re referring to. Is it this one: Filmic, darktable and the quest of the HDR tone mapping - Aurélien PIERRE Engineering

https://github.com/darktable-org/darktable/releases/download/release-2.6.1/darktable-usermanual.pdf, p.97

This module compresses the tonal range of an
image by reproducing the tone and color response
of classic film. Doing so, it will protect the colors
and the contrast in mid-tones, recover the shadows,
and compress bright highlights. It is very suitable
in portrait photography, especially in back-lighting
situations, but will need extra care when details need
to be preserved in highlights (e.g. clouds).

That goes on up to p.101. It’s one of the most (if not the most) documented module in dt.

For some reasons, some parts didn’t made it to the new filmic RGB doc.

EDIT: nope, everything is in the 3.0 user manual too, where it should be : https://darktable.gitlab.io/doc/en/tone_group.html

3 Likes

It’s funny how different people read things differently. This was one of things I read that made me think I was supposed to use filmic since this is a back lit portrait situation. I just wasn’t sure what special care I was supposed to take to preserve the highlights in this case

Backlighting is when you have a (spot) light source behind the subject. Here, you don’t have one light source, the whole sky and ground are acting as the biggest diffused lighting the Earth has ever seen. Extra care is implied in the “compress bright highlights”, since you don’t want to do that in this situation: that extra care would be decompressing them. (Also, for the next time, maybe a polarized filter on the lens could do that for you).

Tone curve food-for-thought…

In rawproc, when I rounded up all my tone operators in one tool, I made a “result curve pane” on a whim. What I do in it is this:

  • Create a 1x1000 synthetic image, and populate it with increasing values from 0.0 to 1.0;
  • Apply the selected tone curve (gamma, reinhard, log-gamma, or filmic) to that synthetic image;
  • Use the values from the synthetic image to plot the curve in the pane.

This bit of feedback has allowed me to do things like compare the operators, and more importantly, use a rather un-intuitive implementation of filmic to good effect. The curve pane updates with each roll of the mouse thumbwheel, allowing me to develop a sense of behavior for each coefficient.

I’ve posted it before, but for convenience here’s a screenshot:
rawproc-filmic

But anyway, here it’s way more than just a tone curve effect. Open the raw file and reduce the exposure, you will see that the snow is solid:

You can massage that the way you want, you won’t get more local contrast with just global operators. You need a serious amount of local contrast operator to bring more texture:

But then, as soon as you raise the lightness (with any kind of power-like/log function), you start loosing those highlights details.

So the way I did it in my first image was, knowing how the local tonemapping works and fails in dt, use its overcooked-HDR-y flaws to enhance the texture. But it works only because the scene is largely monochromatic, and needs to be masked in a hacky way to not touch the darkest areas. Doing the same in a blue and cloudy sky would likely produce dark and desaturated halos in the blue parts around the clouds.

1 Like

I was trying to apply to this XMP file to the original raw, but somehow I cannot succeed… I have placed the xmp file in the same folder as the RAF, and opened the RAF with darktable (3.0.0 under macOS), but only the default basic adjustments are applied and the XMP is ignored.

Any idea? Is there another way to apply an XMP to a given raw in DT?

Yes, but: you need proper lighting for that and IMO it’s a situation that’s solved in pre-production than in post-production.

@aurelienpierre done tremendous amount of documentation, couple hours of video where’s everything explained (well, you can’t cover every corner case, right?) + wrote several tutorials :slight_smile:
However all the tutorials and docs are IMO a “starting point” because in order to understand you’ve got to try.

Nah man - those are a goldmine to those looking for gold. But I bet you could answer with links to specific doc passage or specific time in one of your videos and it might’ve more “give” to it since people will be obliged to look at the docs and go “huh, there’s a documentation” or “huh, there’s well explained video”.

That was one of the things I did - there’s “nothing” there, the differences are minimal and need so much pulling by the ear that they become elves (dunno if that’s an expression :wink: )

Try in ligttable do load sidecar file in history module with replace as operator. That’s the easiest way for me.