Filmic RGB shadows

Todd, not to dive too much into a tangent, but your comment interested me here- can you elaborate a bit on what you mean by " For a shot like yours I often add a linear tone curve and blend it in subtract at 10-25% opacity"? Do you mean set up a mild s-curve in linear mode and then use a subtract lend mode? I’ve not used a curve in this way before so I just tried that as a simple experiment and it didn’t look so good, so I’ve guessing I’m not following exactly what you mean…

You could add some curve or experiment but no…just a straitline slope of 1 . So enable and set blend and reduce the opacity. I find the effect nicer with the tone curve as opposed to the RGB tone curve…I should sort out why they seem so different

Hi guys,
I didn’t give up FILMIC approach and the best kind of universal automatic style for most of my LX100 RAW shots is using the following settings to develop correct dynamic (sure I am also using some sharping, denoise, lens correction etc.). But generally to use some standard settings of Tone EQ for Shadows was a good hint!

Your filmic is ruining the shadows! See orange portion of the plot on the left.

  • Reduce contrast
  • or latitude;
  • or shift the highlight/shadow balance
  • or lower the white relative offset
  • or raise the black point (lower absolute value; less negative)
1 Like

Well, I wasn’t saying ‘filmic is counter-productive and almost there’.

Filmic is the best thing ever and I use it on everything now. Just don’t expect it to be the single one module you need. Use stuff after it to get filmics output to your liking of you need.

I wouldn’t judge what is happening by looking at a plotted chart. If you want more global contrast you crush blacks a bit if you don’t need or want the shadow detail this is fine. Or when you set the black really low in filmic, that means you have a lot of (unusable) blacks, which are fine to push down a bit.

He’s trying to make a ‘auto’ filmic preset… Its bound to be imperfect in at least some cases.

1 Like

It’s not just about clipping: it’s about inversion. The orange part shows that the curve is going downwards. Filmic, by default, maps your selected black point to a non-black value (display tab, target black luminance = 0.0152%). With the curve shown above, some values brighter than black will be pushed even deeper. I don’t know if filmic will clip them to target black (or at least to 0, in case they go negative), and how the modules that come after filmic will react to the situation.

Anyway, if it’s ‘only’ about crushing blacks, then it still means Peter is deceiving himself, since the effective relative black exposure won’t be at what is specified, but where the curve actually lifts off from the x axis. Simply setting a less negative relative black would not lead to shadows being crushed even more, plus the danger of the inversion would be avoided.

I’m also interested in this answer @aurelienpierre ?

I don’t edit like this but when I do it just to test and check I often find that I would easily get the look that I want by just maxing out the contrast slider in Filmic. But then the curve goes all yellow on me and I push values beyond min and max. So how bad is that? What happens then?

I’ve found that doing contrast with tone equalizer on some images is very easy and looks great but on some is very hard and looks like shit. I need more practice with this workflow for sure.

Another thing (and you all will hate me for this). Since Darktable doesn’t have automatic defaults like other software I often use embedded jpeg as a reference photo for sharpness and “perceptual resolution” achieved via contrast. And I find it hard to reproduce the jpeg. I can do worse edits than the jpeg and I can do much better ones, but I cannot create 1:1 edit to jpeg for the life of me. Why I think it’s a good thing to test yourself against jpegs? I think if you can match them in a resonable amount of time, then you can do anything. And I totally get what aurelien said in his last video, but that’s really been bugging me very much. If I can’t reproduce an edit that matches a simple jpeg be it skill wise or software wise, then how can I do anything more or better? I can come close, very close but I just can’t match it to the point of it being indistinguishable. It’s like a little side goal of mine to achieve that :slight_smile:

So is it just a skill problem or are there like a large differences in how camera grades the photo that can never be achieved with other software. Because it’s so annoying that I can’t seem to figure it out what my camera does exactly. I can sorta get the look on one image but then when I apply the same grade to the other image from the same camera it looks totally wrong. It’s annoying as hell.

Basically I have a Canon and Canon is famous for it’s Canon look. I wanna know what they do to the image to achieve that on every photo. What’s the Canon in camera pipeline. There must be one fixed grade that gets applied on every raw and then converted to jpeg.

It’s not clipped since it is far away from 0. the orange part simply indicates that dark shadows are darkened a bit more than even darker shadows or black.
You‘ve to trust your eyes if that‘s too much or not.
Usually it’s about editing an image and not getting a perfect curve …


This is done:

1 Like

Yep that’s why your camera cost so much. Just think about how different all the opinions are about edits coming from capture 1, LR ,DT RT DXO yada yada…every raw processor also uses a slightly different sauce so you are never starting out from the same spot and they may be merging images for Noise reduction etc etc . I could go on and I think there are 3 or 4 very long threads here on this topic. You have to come to the realization that all you have with the raw is the sensor data and what you can make from it …you are not capturing a better jpg. The raw data lets you create something with a better white balance and can give you more room in shadows and highlights and you can save it out in a higher bit depth form but you will often struggle to match the color, tone mapping and noise reduction produced by your camera. Add to that all the in the camera modes, film emulations again etc etc.
If you changes your goals you will be happier. If you like the jpg use it. Maybe even edit it. You can do quite a bit with a decent JPG. Or take your raw and make it your own…

1 Like

Oh thanks!!

Its funny some of the discussions around these points reminds me of the hey day of audio systems. There were amps and big speakers and people argued about what the best amps and speakers were. THey spent thousands of dollars on HI FI equipment capable to 5-35KHz even though the accepted range of the human ear is around 20-20K. To get those specs they paid dearly but they could brag about how great the sound was that they could not hear. Sometimes it seems the same here and maybe its the math based nature of DT but so many people are so concerned with such crazy details only to end up looking at 8 bit images on a tv or the sRGB output to the web or social media. Its not a criticism of anyone or their interest or passion…it just amuses me…:wink:


Good advice, I’ll take it.

To be honest, I can make a good edit in Darktable but I’m really struggling on some images. I’d say the thing I miss the most from Darktable is shaddows, highlights, whites and darks sliders. And I’m struggling with bringing back contrast after filmic.

On some images I can bring back contrast with tone eq right away. But on some images I just can’t seem to be able to do it, and when I’m getting close I end up messing it up and start again. I can also do it using a combination of modules but I wonder if that’s too messy approach? I don’t know what’s the most elegant way of doing it.

And with bringing up shadows and lowering highlights, that’s something that I’m only able to do with exposure module and much of masking. I miss shadows and highlights sliders from Lightroom a lot. But I’d like to learn how to reliably achieve the same with darktable.

Again, I can often do it, but it’s messy and hard. Because I find myself doing it by experimenting while I’d much rather try to have some established ways of doing things. Some repeatable procedure for let’s say bringing up shadows or lowering highlights and that I know exactly what I need to do to achieve the desired result except experimenting much.

And one my say, well just use a curve or something, there are many ways. But in my experience none of them can achieve the results that Lightroom can just by using two sliders. So they must be doing all sorts of things in the background and that’s what I need to figure out. So that I know for sure, if I want to raise shadows a lot but retain all the detail etc, this is the combo and the pipeline that will do that etc.

I’m the one for simple one click solutions, but I need to figure a few things out in Lightroom and Darktable to be able to work more efficiently, not like a blindly.

1 Like

I often use the relight preset in the Tone eq for shadows. Some times I play with it bit often just where it lands is not too bad so I don’t worry about the mask or any thing. The I move the exposure comp to adjust the effect or often as I said its often okay but a bit strong so I just use relight at 50%. Its a pretty quick fix and the results are generally good. I do the same with the tone curve and compress presets in the tone eq. If the image is low key I try one of the tone curve presets in the tone eq and then again just move exposure comp slider if needed and or opacity. If its high key I will use the compress contrast preset and then maybe follow with a second instance of tone eq and now use one of the tone curve presets…I blend this in with local contrast colorbalance dehaze and the contrast eq and those modules are pretty much all I need. I have actually been using the color calibration and not so much the channel mixer but the colorfulness and brightness sliders…you might want to try them…esp brightness…darkening blue bumps the blacks and green is related to luminance so you can play with those…I guess that with some masking and blend modes from time to time and that is about as far as I stray. Now that I think of it I do use color zones quite a lot for small fixes. If I have say a rock face with a cast I don’t like. I don’t mess with WB i just auto range pick from the saturation tab of the CZ module with a shift to create an auto desaturate. It often does a great job as is and often without the need to mask. I do the same with skin sometimes to remove color cast. There are a few technical modules that I use. I don’t have an extensive lens correction but even for my smartphone the vignette correction is amazing using lens corrections. If you have shadows and you don’t use the lens corrections you should try because this could be missing boost to the shadows…anyway that is a few random thoughts…keep plugging…believe me I started exactly like that. I was obsessed and could not figure out why I could not get a better result…over time I embraced what a raw file was and then changed my approach to use the inherent strength that it offered…here’s proof…I had the same question…Attempts to get a raw edit that matches the jpeg....

1 Like

Color Balance > Shadows part > Factor slider. Decrease it to darken shadows.

1 Like

I do everything pretty much the same as you, basically use everything to get something like what you pictured in your head. But I’m becoming obsessed in doing it using as little modules as possible. Although one difference is, I rarely use tone eq as a tone curve without masking and I might need to experiment more with that.

But for shadows, highlights, whites, blacks tone equalizer is just not an equivalent. I just can’t get the similar results. There are so many variables, so either I’m stupid or I’m not yet skilled enough or tone eq can’t do it. I’ve tried rbg curves, zones, color balance (that one I might want to revisit tho).

Raising shadows with color balance is also weird, it affects the whole image and desaturates it (or doesn’t fix chroma and then it appears like it’s covered by a half opacity pure white color). I can get something with different blending modes but it’s still not really that great.

I’ll probably dedicate a few days and really test everything and every combination I can to see if I can figure something out. There must be some kind of module + blending mode combo to achieve a great result. If I can’t I’ll give up and toss it here for the fine and smarter folks to figure it out.

the new color balance might be onto something! When shadows are lifted, they aren’t desaturated, only problem is It seems to raise the highlights and the midtones and I loose all the detail there but I still need to do more testing, I’ve literally just opened it and pulled the shadows perceptual purity grading up so I’m yet to understand and test it properly.

Edit: Oh man, aurelien is really onto something with this new module. It’s still not 100% there yet but wow, I just had a big rant in previous posts and now I see this exact thing being addressed. Colour me impressed!

Is this the one from version 3.4 or are you running master?

This is just the nature of life :stuck_out_tongue:

1 Like

Yeah, the master, the color balance RGB module. It’s much much better but it can obliterate any details pretty quickly. So it’s kinda there but not yet.