I hope this doesn’t come across as a silly opening question, but as a long time lurker and somebody who has been playing with Darktable for 18 months, I remain unsure of the rationale for using filmic/filmic RGB.
I understand the arguments against using a base curve and I comprehend the thinking behind preferring RGB over Lab. My own experiences suggest that some form of curve is usually required for decent images.
What I cannot however fathom, is whether people are electing to use filmic because it is more consistent/numerical and involves moving sliders as opposed to randomly dragging a curve around, or are they doing it because they prefer images with less color in the shadows and highlights, which as understand it was the author’s original intention.
I get the impression that it is the former, and that filmic is essentially being used as a curve tool, but I’d be interested hear what other people think.
Well… I think there are effects that you can achieve with filmic that you’d need to use masks for if you tried to achieve them with other filters/modules. And you’d probably also need several instances. E.g. you can reconstruct highlights with the exposure module as well, but you would want to exclude not so bright parts from darkening.
But I am not quite sure about this, I am not a dev.
That’s not true, in my humble opinion.
There’s a slider in the look tab (shadows/highlights) that I always tweak, because it allows me to recover the colors I want. It works specially well when I want to recover deep blues in the sky (but also for any other strong color). It is 12% by default, and usually I bring it down to zero, and sometimes even to negative values (for strong colors on highlights). And I must say it never failed me.
DISCLAIMER: My edits are usually more on the saturated side.
The intention of filmic is to transfer data of high dynamic range (sensor data) to a representation with low dynamic range (e.g. monitor). This is done in a structured way and should represents the original sensor data in a “natural” way. Less color in shadows and highlights is just what is to be expected ( because pure white and pure black have no color information).
After using filmic to generate a decent and “natural” representation of the original sensor data, you can still use other tools and go “artistic”.
Base curve is actually (IMHO) impossible to manually adjust. Filmic RGB gives you FULL control. That + linear editing that comes with it means you have way more room for adjustments. And that “less color in in shadows/highlights” is blown out of proportion. I actually get better colours thanks to better edits overall.
Also - filmic + tone equalizer is godsend for wider dynamic range that would take ages to normally edit.
I find filmic particularly useful for high dynamic range images. For the majority of images, however, I find the base curve or a LUT similarly effective, and less hassle.
So for me, personally, Filmic is a specialty tool for difficult situations. I recon that this is a result of filmic’s linear color space, which allows HDR processing, whereas the other modules only work for small dynamic ranges. (I don’t know if that’s technically true, though).
I was of a similar mind until I re-examined some of my earlier edits. Where I had used the base curve, I found that blue skies had taken on hue shifts that would take time to correct or otherwise live with. The same skies retained their color in Filmic RGB, although sometimes they were desaturated - a condition far easier to correct that adjust the hue.
I think the learning curve challenge with Filmic is that you’re not after a set of steps, or even a common end state, but rather you’re setting boundary conditions on exposure and contrast that enable the module to create a response.
But it becomes a very straight forward exercise once you know the workflow and have a sense of what to expect when finished.
That’s true. I had forgotten about that, as I had switched from base curves to camera-appropriate LUTs a while ago. (Fuji Pro Neg or Astia usually)
I generally use LUTs as my default starting point, and only use filmic in a minority of cases. After some experimentation, I simply found my manual filmic based workflows to essentially converge on my chosen LUT anyway, so it simply made sense to go that way by default.
I was sceptical at first and now use it as pert of my standard styles.
basecurve can do the job, but it’s not easy to adjust manually (as noted by Johnny-Bit). So I would make an adjustment in exposure to fill the spectrum to the right edge, apply my standard basecurve, and fix the rest in tonecurve. For what it’s worth, when I had barely to touch tonecurve, the results were very good (technically), but sometimes the best photos are not the ones which are technically spot-on.
filmic RGB being parametric, I can easily fiddle it a bit to take account of the lighting. I then still use tonecurve, but usually for much smaller corrections, and the final result seems better.
Although tonecurve is not supposed to play well with filmic RGB, I’ve not had any problems. There may be if you still have a version from before the power-norm bug was fixed, although in B&W this really didn’t matter.
I still don’t use tone equaliser. I can see the advantages if you want to make a sort of global lighting adjustment. Maybe combined with masks it could do interesting things, but I’m very sensitive to inconsistencies of ligh across the frame. Conversely, tonecurve allows me to make relatively sudden changes in contrast to pick out, for eg, features of a face in shade. I needed to be convinced to invest the time to learn to like filmic RGB, it may happen that I’ll take up tone equaliser one day. These thoughts remain valid only until I close this window
PS: basecurve in 3.0.0 and beyond is really a different beast to pre-3.0… it used to be applied near the start of the pixel pipe, with most other changes made after. In 3.0.0, it was moved to the end of the pixel pipe, in roughly the same place as filmic rgb. So the big technical change was moving most of the manipulation tools prior to the tone mapping applied in either basecurve or filmic RGB.
Bringing it down increases saturation on the horizon. Maybe not the best example, but you can see it changing. I’ve worked on images where the change was more prominent.
EDIT: While preparing this image for this post, I noticed that increasing White in the scene tab also increases saturation, but at the expense of loss of contrast. I particularly prefer to set White as bright as possible and then recover colors with the shadows/highlights slider if that is the case
I have finally come to terms with Filmic (RGB). It has been a long process but my end product has improved and my processing time diminished. Here is how I now do my processing.
The pixelpipeline is a remarkable help. I can preset most of my processing in the knowledge that dt will ensure that the correct order will be maintained. With that in mind I preset standard settings for ‘local contrast’, ‘LUT 3D’, ‘sharpening’, ‘raw+profiled denoise’ and ‘lens correction’. All that is really left to do is to apply my Filmic RGB.
What I find in the latest versions of Filmic is that the spot picker for middle-grey-luminance does a wonderful job of finding a very good starting point for setting this module. Yes, it may need a bit of final tweaking along with the black and white sliders but on the whole I am generally 90% home with this single click.
Of course, a decent exposure helps immeasurably and yes I do tweak other settings if the contrast is tricky but this general approach does get most of my processing done in a very satisfactory way and in short order.
One thing more that I do notice … the look of my images has changed with this system. I believe it is a more ‘natural’ look with less signs of processing.
I still perform local adjustments on special images that deserve the extra treatment but basic pprocessing is now a ‘breeze’!
Thank you for your quick response. You are right, the saturation along the horizon has been improved a little bit by changing the shadow/highlights setting………
If you pull up the display on the current DT 3.4 if you have it and show the ev display not the curve you will see better what happens…That slider shift the latitude region ie the saturation protected area and so shifting it left ie the slider shift the latitude tone mapped area into the highlights…it is very easy to see in that display…I am not on a PC with DT so I can’t put a screen shot…
EDIT pulled this from the blog…using this display play with the sliders…it really is the best way IMO to see what filmic is doing …when you move the slider you will see the latitude shift…
There is a lot to be gained in DT by moving modules around filmic. Boris Hadjukovic does it quite effectively. Often moving instances exposure, colorbalance or local contrast after filmic. As filmic comes later in the pipeline it can be better to introduce color and contrast changes after letting it do its work…just an observation…
The color preservation mode used in filmic has a strong effect…I think it defaults to power norm…it warrants checking from none through all the options…
