See Filmic v6 by aurelienpierre · Pull Request #10976 · darktable-org/darktable · GitHub. Many things are harder in scene-referred, not because they are actually harder, but because display-referred hid them under the carpet and distracted us. Also, display-referred had 30 years of heads-up, so of course they had time to make pretty UI and easy stuff. It’s kind of unfair to compare it to something 3 years old that is still ignored by the industry (although : Gradient interpolation in Photoshop)
2 Likes
Was this suggestion not in the manual for some time??
Note: filmic rgb cannot be set with entirely neutral parameters (resulting in a “no-operation”) – as soon as the module is enabled, the image is always at least slightly affected. You can, however, come close to neutral with the following settings:
- in the look tab, set contrast to 1.0, latitude to 99 and mid-tones saturation to 0 ,
- in the options tab, set contrast in shadows and in highlights to soft.
In this configuration, filmic will only perform a logarithmic tone mapping between the bounds set in the scene tab.
That’s great but you forgot something… output. The point of filmic is to introduce a logic suitable for any display (HDR or SDR). Of course, as long as everything is SDR, tricks like that work. Which doesn’t mean they are future-proof. And the point was, precisely, to move on to HDR gradually.
2 Likes
Well, no, if it is separate, the user can decide if we wants to each part or try other kind of curve/transformation.
Another advantage is being able to control masks and other parameters separate.
But I understand what you say that they are quite interrelated among them and not easy/desirable to separate.
May be the way is a kind of “submodules” than and be able to select among alternatives.
Anyway having interface separate, does not mean neccesarily processing each of them separatly, may be you combine the equations and curves befor applying them in one step (easy with linear transform, not so easy with nonlinear).
Again that can be quite more difficult and much more work. So the way might be implementing alternatives inside the filmic module as “plugable” funtions and interface tabs.
Thanks for the clarification.
I like filmic a lot … when I don’t need to touch anything (most of the times, when the main subject is in mid tones or no need to fight the highlights).
Yes that’s been there for a while but I think the intention might have been more in case people want no-op as a starting point.
Does it really matter? Don’t you have to tweak it anyway? Your sdr and hdr edits cant be one. Tweaks are required. I mean the same look don’t even work for two print sizes.
For that use case turning of one look module and on another seems more flexible?
Yes, that is true. Leading the way and making things differente and with a new approach is not easy and there are difficulties.
We have to thank for the effort (and for the great results) and hope that that aspect that keep being more difficult will improve with time as they are taken to light and a way of improvement shown.
Sometimes it is just a matter of being used to other workflow , there are tons and tons of tutorias on how to do things in PS or LR, with differente levels of technical explanations, not so much using DT and the scene workflow and many of them too technical for many users (it could not be other way).
There is time since I heard of floating point 32 bits linear processing of images… but few software has gone deep in processing in that model of workflow.
I have no doubt it is a better model.
No op is no good because the log mapping is gold. Had completely missed that this discovery is old hat and was part of the manual.
1 Like
You mean you’re not going to break into my house at night and delete the basecurve module code off my computer like the panicked threads seem to think is going to happen?
No one listens to me, probably for the better, my life is a disaster.
@nosle
I’ve been using the flat filmic + curves method for years, although I use rgb curves in lightness blend mode so I can tweak each channel independently. I wouldn’t bother doing this for everyday images, but it’s worth comparing to filmic for portfolio images. Were told the drawback is that it’s not a suitable workflow for hdr. I don’t bother about this - if I cranked screen brightness up 10x my eyes would get sore, so I find it hard to believe hdr will catch on in a big way. Unless there is more to it I don’t understand. Anyone making photos for print certainly doesn’t have to worry about hdr.
That being said, filmic improves with each rendition, and increasingly produces results as good or better than my custom rgb curves. The New Euclidean ratio, plus ‘safe’ bounds, has helped a lot.
1 Like
I am a hobbyist photographer who started shooting raw in mid-2021, and started gradually learning Darktable for this purpose. Because of this, practically I have almost no prior experience with the display-referred workflow in any depth. I used GIMP to edit my OOC JPEGs before (rotate, crop, tweak the levels, done).
In Darktable, I started with scene-referred, and it made sense intuitively from day 0. I am surprised that what is called “display-referred” actually exists for serious software written in the last decade or so. It makes no sense — transformed spaces may be reasonable for representation (eg to save memory, or improve accuracy), but surely you want to perform operations in a linear space as long as possible, otherwise 90% of programmer time will be spent compensating for artifacts of the transformations.
For me it is not about SDR vs HDR, but being able to think about the pipeline as a series of transformations I can understand and reason about.
9 Likes
Yep, nothing so hard as to unlearn acquired habits… Once I manages to ignore those reflex reactions, I also found the whole linear pipeline easier and faster than the old display-based workflow.
I think part of the difficulty for some is that editors like the GIMP work in strict application order, so you have to do some operations in a specific order. That’s not the case in darktable, and getting used to that also took me some time.
Inertia.
You need someone in your team who understands and is capable of implementing the scene-referred workflow. And you have to convince your users it’s worth the investment to retrain (probably a much harder hurdle for the professional players). “If it ain’t broken, don’t fix it” may also play a role here (it might have been broken, but if you don’t notice that…)
3 Likes
I always found (with display-referred and using tone curves) that there was a lot of back and forth trying to juggle modules because I fixed one thing and broke something else as a byproduct (colours and contrast, mostly). It feels like I do a lot less of that now.
1 Like
Reading the comments on this thread is hilarious frustrating.
The developers added new functionality to DT. They explained the logic behind the new modules and have YT videos on how to use them. There are a ton of other YT videos and playraw images that you can use to learn besides just the manual.
I read some of the post here that some don’t like the new functionality. Or don’t agree with it. Or think it is wrong. Or the overexposed sky/clouds dont look how they want it. Or whatever. And it appears that the post comes from those that have not spend the time trying to learn the new process or even read the manual.
But no one is forcing the use of the new scene process. The other modules are there and you can still do the display process if you want to. You can do a hybrid approach to get the image you want.
4 Likes
Do they exist though ? It’s hard to tell exactly what closed source products do but I doubt they start their processing pipeline with a tone curve or some crude colour transformation. E.g. from Capture One doc :
Capture One works in a very large color space, similar to that captured by camera sensors. A large color space ensures that little clipping of the color data can occur. Clipping is the loss of image information in a region of an image. Clipping appears when one or more color values are larger than the histogram (color space of the output file).
At the end of the workflow, the RAW data has to be processed to pixel-based image files, in defined color spaces. These spaces are smaller than the internal color space used by Capture One. When processing, some color data will be discarded. That is why it is crucial to perform color corrections and optimizations before processing images in a smaller color space.
I think there is more to it than just a “large” color space that is a gamut and transfer curve combo and you can have issues like this
and these
2 Likes
Are you sure of that?
Most of the transformation related to luminosity and even colors, are not noticeble in hightlihts and exagerated in darktones if done linearly.
So what you call “work in a linear workflow” really means that the software has to apply strong non linear transformations to your linear data to show what it shows.
A workflow with a gamma space make somethings easier to get, and has a behaviour more similar to human perception.
To get what DT gets now working in linear space it has to do a lot of work and do things in different way, but it gest better results, less problems with halos…
Anyway the used space is not an “untransformed space” either.
But no doubt for me that the current scene referred linear, non bounded space is the future.
Some difficulties will improve with time, traditional software took time to improve too.
1 Like
Almost. A few modules have been removed for new edits, and a few are marked as deprecated. One of those was shadows and highlights which was rather popular. So some will be forced to change their habits and learn new tools (or use tricks to get around the changes)
1 Like
I’ve been using filmic rgb in tandem with rgb curve (placed after/above filmic) for all my edits. You can see it in all of my play raws, I believe. However, just recently I felt I’ve got the hang of using them together to produce the results I’m after.
When you say this, @anon41087856:
…does that only apply to base curve or does it apply to the rgb curve and tone curve module too?
I’m asking because I’ve avoided base curve like the plague because it does not have the cubic spline interpolation method. monotonic (only method available in base curve) frustrates the hell out of me. Both tone curve and rgb curve has a way to switch to cubic spline.
Also, would there be any ill effects to placing drawn mask instances of the base curve (or tone curve or rgb curve) module before filmic rgb in a scene referred workflow? I just love local adjustments with good ol’ curves.
I’ve also grown finally grown to love filmic rgb and its abilities to gracefully transition to blown out highlights with the help of blooming. 
