Thanks. playing the the default style is interesting. Out of the box it produces quite a useful result but in this case with the darks too dark and the oranges a bit over saturated. Could be the camera settings and such. I use a Sony a6500 these days but have photos from a variety of cameras that I need to set up styles and defaults for.
I guess I need to look at the order of operations in DT too. because it I turn off your settings for contrast equaliser, colour balance, and local contrast (just leaving filmic RGB) I am more or less at the best that I could do with just filmic RGB too.
Maintain full dynamic range is not possible, it’s not displayable on current output devices. Filmic applies curve that protects midtones and compress the high and low parts of the DR.
You can see this video from filmic developer where the compression is explained
In any case in high dr image you will have to do a compromise.
There are some trics which may help to recover some part afterwards, e.g. idea from Boris Hajdukovic
Here’s a take using Pierre’s Filmic-based style along with an exposure and black level adjustment. To get more details in the highlights, I applied an rgb curve, applied after filmic, to increase contrast in the birds’ breasts.
The XMP is from the current Git master (commit 9fc00e3f), so may not be usable in your version. So here’s a screenshot of the rgb curve settings and placement.
Your dilemma is one of the reasons I’m not too enamored with regard to attempts to abstract the filmic curve. Most tone modifiers can to an extent be described with a curve graph, which IMHO is the only intuitive way to assess the impact of a tone operator.
This screenshot is not darktable, but it shows a filmic class curve modified to address your specific image, along with a curve plot of the actual transfer function:
This filmic operator is the original equation laid out by Duiker, controlled with the coefficients A, B, C, and D. These coefficients have decidedly non-intuitive effect on the curve; suffice to say what I’ve dialed in is substantially different from the defaults, in order to flatten the log-style shape while retaining that little “ski-ramp” shape of the toe. You can see this easily in the curve plot in the lower left.
Thing is, you can do the same thing with a regular control-point curve. The point is, the filmic curve is designed to lift image values while preserving a film-like crush of the lowest tones. That lift is apparently not friendly to penguin chests. It took a rather aggressive modification of the defaults to address the need for retaining tonality in the highest parts of the data.
I’m not dissing darktable’s filmic, it’s a very powerful and appropriate tool for tone manipulation. I just want to point out that the curve it inflicts is an insightful thing to consider regarding the tool’s effect.
One of the things that I struggle with in it’s implementation is that it, as you say “inflicts” a curve immediately on activation. I would prefer if it, and all other tools, behaved more as they do in other apps in that on inception they start with the image unaltered and then move from that as you start to tweak.
Sorry I couldn’t reply to everybody. I ran out of reply credits! Anyway thanks to everybody for your inputs. I have made some good progress. Here is an edit of another photo on the same theme, from the same file. Basic edit flow was exposure, Filmic RGB, local contrast with multiply blend, and iterate a couple of times. I’ve learnt a lot about using styles to and will work on a basic style for each of the cameras used in my library.
Hi thanks for your comments. The halo was caused by the tone equaliser module I think, but maybe by local contast, I remember having to back off. The lens is unsupported and I think it would have been a Tamron 11 - 18mm that I bought for creative fun knowing that it would likely have issue. I don’t have that camera (Canon 7d) or lens anymore as they succumbed to tropical mildew at some point. Will look into the chromatic aberration module. Thanks again.
The chromatic aberration module worked quite well. ‘defringe’ did nothing.
The halo and funny fringe nearby on the penguin’s down is caused by the “tone equaliser”. I was able to get rid of most of the halo and fringe by cranking up the ‘filter diffusion’, ‘smoothing diameter’, and ‘edges refinement/feathering’ sliders.
Ooh, good question. In most of the tools I implement in rawproc, I hard-code a neutral default, but I didn’t with filmic. In the research I did to implement it, it was always presented as a modifier from the start, with values in the coefficients designed to do a certain off-axis shift of the curve. Conversely, the thing we think of as “gamma”, is really a simple exponential function, which if you set the exponent to 1.0. results in such a neutral default.
In rawproc, each tool has a checkbox that, if checked, tells the software to route that tool’s result image to the display. To regard filmic’s starting point, I’d just click the previous tool’s checkbox.
But, now that you’ve brought it up, in a spare moment I’ll play with my filmic tool, see if I can dial-in a set of coefficients that “straighten-out” the transfer function so X=Y. I don’t think I’ll be successful, based on the structure of the equation, but what else is there to do in solitary confinement…
Chris welcome…as you often find you need to go to the color balance module for adding back some saturations I also will use the picker to select the contrast fulcrum and use contrast in the CB module …sometimes that is all you need…obviously image dependent…
