Or amplify the noise, for a well exposed shot there aren’t any pratical reason to use log gamma.
Log really shining when a photo is under exposed, but if the noise is amplified too much it could be due to a low dynamic range from the camera or that it was choosen an irrealistic mid gray
You have other algorithms to deal with the noise. It’s always a matter of trade-off. Every transfer function that raises the shadows will increase the noise (even a simple exposure lift), so this is not specific to the log.
What is log gamma ?
Must say that your results make a healthy start to the development process. I must admit that I cropped prior to any processing which in turn may have eliminated some extreme levels.
I am sure that you have a better ‘feel’ for this module than most of us; the logic, despite the manual, is somewhat alien to me at this time … and yes I did RTFM (printed it too).
Hi there,
darktable 2.6rc1 should bring a more stable framework to find the best settings in Filmic, with a much clearer curve to read and no more weird jumps between close parameters values. Merry Christmas !
The filmic S curve has improved a lot with the 2.6 release. Control points help to understand what happens.
But the curve shape depends on spline curve calculation which sometime doesn’t produce expected toe and shoulder. However this is usally not an issue.
I’ve more difficulty to control saturation especially on landscape. Filmic tends to desaturate highlights. And depending on contrast we get easily shadows oversaturated and washed highlights.
Maybe a specific control of saturation for shadows, midtones and highlights would turn this easier.
This is the kind of image which is problematic for me (dt snapshot, tonecurve based, not too far from the scene I have seen), I cannot get anything close to that using filmic. Any tip ?
I would agree that there appears to be a significant improvement in the function as of late. I routinely ‘test’ the advantages of filmic against my more conventional multi-step approach.
With a number of my images I find that filmic responds well to the ‘almost automatic’ one-click operation and then needs only fine tuning. However, that said, for reasons that I do not understand on many occasions filmic simply cannot provide any sort of reasonable starting point and working with it is simply a waste of my time.
I question if filmic is a ‘better’ mousetrap or simply an alternate route. Whichever path that I take with my RAW processing I am convinced that filmic is not saving me time in my work and I find that it causes a loss in creative flexibility.
I will continue to play with it and watch future developments with interest.
I think it is another tool in a fairly robust tool bag.
It’d be helpful if you explain the “loss of creative flexibility” comment.
A bit dull, so far, but in the right direction, I think?
Have fun!
Claes in Lund, Sweden
– filmic experimenter –
The main thing I would say about filmic is to know what it is and isn’t. If you are looking for a tool that “would solve all of your problems” and “magically fit with every other tool in your workflow”, this isn’t it.
I believe that my ‘normal’ multi-step process provides more flexibility than that of filmic where the single module tries to encompass much of the functionality of several modules (such as: contrast, saturation, levels, exposure, tone-curve).
I have found through numerous tests on my own data that, for me, allows greater flexibility.
Another thing that I notice is that attempting to use other modules after an initial usage of filmic may require added time fiddling-about further with the filmic settings.
I am still finding images where the ‘auto-tune’ initiates impossible setting … for instance with a DR that is beyond all reason.
Yes, I do agree that it is another tool in our box … an an interesting one at that. Although it was, I am sure, not designed to be a one-click solution to the RAW process, it still strikes me as having that feeling.
This is what I was able to accomplish with filmic only.
PS @anon41087856 When lowering balance shadows-highlights, there is a point at which the blue in the shadows of the histogram leaps to the left, and then the shadows, which were desaturating, suddenly becomes very saturated.
To go a bit this way, I think filmic would be easier to use if we could gradually activate the features: 1, log curve, 2. S curve, 3. saturation(s)… (EDIT) and see the corresponding effect.
With filmic, and unbreak as well, black color picker (and related automation) is very dependent on the noise which can create artefact values. But it is quite easy to fix this changing the black point (the histogram gives good indication for that; to be more accurate you can use the color picker (min)).
I get also better result when I let some room on both sides (shadows and highlights) in exposure module (no clipping).
I agree !
The kind of image which seems to be difficult to treat with filmic, is when the details I want to get are in shadows and highlights more than in midtones. The S curve tends to compress them on both sides, giving the opposite effect.
In that case I think I should compress the midtones instead and get (or keep) contrast in shadows and highlights, what I did with tonecurve:
It really depends on your learning and communication style.
The tone curve may be easier to grasp because it is more visual and direct than filmic, which is why I completely understood where @gadolf was coming from when he asked if we could adjust the curve of the filmic module directly.
Agreed. And I think the tonecurve spline curve is probably not the best tool to mimic and control a real filmic curve with toe and shoulder …
I look upon images as being the sum of their numerous parts. Extracting, for me, the appropriate amount of shadow and highlight detail is critical in how I wish to see a final image. The mid-tones may dominate an image but the S&H are the icing on the cake.
I can agree on this. Filmic works very well for me to extract shadows from dark along with midtones. I struggle to get good highlights (detail and especially saturation).
In fact unbreak works pretty well also to extract shadows (thanks to log curve). But the correction to be done on top of it (tonecurve or color balance) is usually severe. I’ve work a bit to build a sigmoidal curve as a second step in unbreak to decrease the afterwards correction (https://github.com/phweyland/darktable/tree/ub-sig). But that doesn’t help neither to get (to keep) both shadows and highlights.
That is a nasty one, but you need to decompose what’s going on : your scene is not HDR (I get about 8 EV of dyynamic range) because of the haze that diffuses light and wash your blacks. So, filmic will not be able to give you the contrast you expect all alone, you need the dehaze module too.
Filmic is not a creative module, it a sort of view-transform, trying to squeeze the dynamic range of the camera into the dynamic range of the destination medium (and uses the film behaviour as a reference). Its benefit is to keep the scene-referred ratios untouched during non-linear operations, that would normally destroy them (that is, with chroma preservation on), and let you deal with them manually (although that can be trecky sometimes)
It desaturates highlights and shadows outside of the latitude, so you can manage where it desaturates with the balance shadows/highlights, by sliding the latitude. In your reference image, I find the bright highlights have an unpleasant piss yellow, because the red is clipped, so the only solution to me is to let them degrade toward pure white.
What did they do in the darkroom ? Dodging and burning. You can always mask the highlights with a parametric blending and the mask feathering.
So here is what I get:
Not masking (filmic + dehaze + a bit of contrast in color balance + local contrast):
Same but masking highlights (this piss color is really disturbing):
