Filmic maps the scene dynamic range to the display dynamic range. It’s not artistic, it’s technical : “white” on the scene may be 1600 times brighter than “white” on your screen, so we need a not-too-destructive way to convert the dynamic range (and gamut) between media.
Since this conversion usually involves compressing tones quite a lot (aka flattening the contrast), the filmic contrast slider is provided for damage control, to restore contrast around mid-tones. Think of it as a mitigation operation. It’s not artistic.
Color balance RGB contrast is meant only as a shortcut to quickly change the contrast of masked parts when used with masks (for example : enhance the foreground). It should not be used globally as it changes the apparent dynamic range of the picture and triggers circular editings when filmic is already set.
Thanks a lot!
I’m not an artistic postprocessorer, I love a realistic look as I remember the scene. Many pictures in the web are oversaturated, oversharpened, over-all-ed. My intention was to compensate filmic. Sometimes, the contrast value of 1.35 in filmic is little too low. Filmic follows in the pipe after “color balance rgb”, therefore it is logical to do this compensation in filmic. Somewhat stupid, my question.
What about restoring contrast to highlights? Is there anything built into filmic that should do that or is it just other modules like local contrast? I’ve recently discovered that I can retain a lot more detail in highlights, especially clouds, by turning off filmic and just using the tone equalizer for all tonal edits. But this can be more time consuming and I miss out on the other nice features of filmic. If I’m not mistaken, filmic always compresses the highlights, even when I’ve exposed for the highlights and don’t need them to be compressed. Any recommendations?
Of course. The desaturation is a rough gamut handling.
But that’s by design… The output dynamic range is a finite resource. When the contrast is set > 1 (meaning the slope of the central part of the S curve becomes stiff), it means you expand the DR allocated to midtones. If you expand midtones in a finite range… obviously, you compress highlights and shadows. Hence the loss of local contrast near white and black.
So, either you set contrast close to 1 in filmic, or you dodge and burn the old-fashion way (masked exposure or tone EQ if regions are not sharply defined), but there is no magic trick that will allow to expand both midtones and highlights contrast at once in a finite dynamic range.
On another note, I think people overdo local contrast in highlights, especially in clouds. Look at any sky with your bare eyes… since you are half-blinded by the light intensity, you don’t see much anything in there anyway. More on that in a another post…
Thanks for the explanation.
I mainly do landscape photography and clouds are often the primary subject in the frame. At certain times of the day and with certain clouds (cirrus), they are not blinding to look at and have amazing detail. I expose to prevent highlights from clipping, so all this detail is captured by the sensor.
If I want to give most of the dynamic range to the highlights, effectively compressing shadows and some midtones only, is there a way to effectively “turn off” the compression done to highlights? If I’m not mistaken, setting the contrast to 1 in filmic still compresses the highlights. I’m fine with not using filmic at all for certain shots, but wondered if I was missing something.
You might be able to do something using the “shadows/highlight balance” control in the look tab and/or the “contrast in highlights” control in the options tab.
You could also try setting your white relative exposure so that those details aren’t right up against the edge of the histogram.
I doubt its coming because clouds have incredible detail and depth when viewed with your eyes. Its part of basically all cultures to admire their shapes and detail.
Of course its made possible by the constant exposure adjustments of the eye and the image processing of our brains. Photography has to deal with the mismatch between a static image and the process of our vision and experience.
Ever went to a museum ? Because painters had all they wanted to paint super dramatic clouds, and yet they didn’t. I did a study at the Rijksmuseum in Amsterdam, and it turns out clouds are really more muted in paintings than in photography. That tells me the oversharpening crowd fucked the game in-between.
Sometimes, I see super dramatic clouds in real life. And then I realize my sunglasses have a polarizing filter. When I remove them, it’s just cottage cheese in blue sky.
I’ve worked on the design of several. To get to that position you have to have visited some. (understatement)
I’m sorry this is just absurd. Some did some didn’t Constable springs to mind but there are so many and anyone just thinking briefly should remember one, by a famous artist or just a print in a toilet. I think you are getting unnecessarily stuck in your head forgetting things you’ve actually experienced.
The fact that most landscape photography on the web is overcooked is a different issue.
I know that Mantegna is an extreme (not unique though) example and he had an different background than we have today. But what I want to say here is that we have different references for depicting clouds and that the beauty of darktable and modules you made is that they give us greater possibility to imitate them as we like.
What you are talking about is not the Mantegna but the picture which I found on the internet and reluctantly posted here just for your reference, aren’t you? It doesn’t represent well what we can see in Paris, let alone the original appearance of the painting in the 15th century which could not be so “dull”. I can show you a more colorful image of the painting although it is still not a good reproduction. You may also refer to the version of the same subject in the Kunsthistorisches Museum in Vienna which preserves the vivid colors from the Renaissance era better and shows you Mantegna’s unique depiction of clouds as well.
You can use masking to achieve contrast in highlights. For one example, assuming ETTR, create new instance of exposure module and lower exposure. Then in parametric mask, luminance channel, move the black sliders right so the shadows and midtones are not being affected. Then, to create contrast, move the top white slider left towards the top black slider. So the darker parts of the highlights have exposure reduced more than the lighter parts - thus creating contrast. If you want this to only effect a certain portion of the image (eg. the sky), then you can then draw a shape mask around the sky.
For me the need for damage control (this coming from the „inventor“ of the module is making this even worse) means that something is fundamentally wrong with the whole concept of the filmic module… I don’t think that a module should “damage” the contrast that late in the pipeline that it needs major repair work - it means that the module basically is destroying a lot of decisions the photographer has made beforehand, either in camera or in basic development settings.
Contrast and contrast distribution regarding our output media is a thing we photographers - unlike snapshooters - have always in mind and decide very early on in the photographic process by deciding our exposure and the way we light or scene. The larger RAW capabilities protect us from mistakes we make to a certain degree but compressing or tone mapping the captured data late in the edit is completely the wrong way to approach this unless you want to have a HDR image (yuck, I hate HDR look thoroughly) and only then this module behaviour IMHO would be acceptable.
To the OP: switch from scene referenced to display referenced mode and forget about filmic module. The whole scene referenced malarkey is a misconception that only benefits camera users (they aren’t photographers in my regard) who care more about technically not losing signal rather than creating a result according their vision/perception of a scene. It’s the difference between technically measuring something and visualizing it so that every nuance is visible as needed for documentary purposes as compared to capturing a scene in an artistic way.
If I were a forensic detective I may need the former but as a photographer I must have the latter, Darktable will probably be not viable for that in the long run since many of the needed filters fr the display referenced workflow have been sacrificed on the altar of technical correctness and are deprecated…
You’ve already said this several times in other places, but you have only made an emotional argument that seems to stem from the fact that you do not understand the new workflow. That is fine and you should use whatever tool works for you, but showing up to continually malign something you clearly do not understand is not cool and you should stop. Stop until you can provide a logical and technical argument to prove your point.
Again you’re confusing position in the pipeline with the order in which you turn the module on. Filmic is the third module one uses, so plenty of time to shape the look one wants.
The pipe is largely nondestructive, so no, nothing is destroyed.
This is true if you’re shooting film or jpegs, but not if you’re shooting raw. The general recommendation to use scene referred is to ETTR.
Good thing you don’t get to define who is and isn’t a photographer. Its a wonder you can see anything with your head that far up your own posterior.
I don’t necessarily think it’s a question of oversharpening or even realism, but rather a creative choice. Maybe those painters didn’t want clouds to distract from their primary subject. Their clouds were probably background elements.
But sometimes as photographers we want the clouds to be the primary focus.
If this Filmic malarkey is the third module one uses - what about the „scene referenced“ workflow, after that damaging module you no longer are in scene referenced mode, you are in a film simulation tone mapped mode. So either this whole „scene referenced“ thing is just a buzzword or the module is late in the pipeline but must be configured and fixed early (idiotic and unintuitive workflow at the very least!) because otherwise you keep running into the endless back and forth of the edit requiring changes to the contrast which requires changes to the filmic settings which requires changes to the contrast because the „damage limitation“ is insufficient. A module that requires a „damage limitation“ setting IMHO is fataly flawed in concept and the developer should go back to the drawingboard and fix the concept instead of tieing a big unsightly dressing on the open wound!
I don‘t want to have to decide on a tone map that early in any workflow, I don‘t care about tone maps - I make decisions on the target dynamic range depending on my target media, whether it is a traditional print , an offset print, a web presentation or a HDR display, often several which is why all of these must have a common edit, because my subject will always be smack bang in the available dynamic range of the most limited one of the ones I currently need (often large format metal prints which have 5-6 EV dynamic range with some discontinuities in the highlights). That will also require different sharpening (so sharpening is an export setting) and different handling of shadows, color ranges and highlight handling. This is easily handled in the „old“ workflow but impossible in the new one which enforces decisions much too early to be viable!
Darktable now is an unintuitive joke of an ego project since the switch to the „must not drop technically recorded infos at all cost, even if it looks shite“ mode…
