Yes. That’s why I emphasized that this has been taken to an extreme. I know how to work it around. I’m trying to point out that it behaves wrong.
Well it now behaves differently again in v6…and does not look like that… also what happens with the norms is that in extreme highlights channels that are clipped or near clipped in v5 don’t get corrected properly so you get essentially and exaggeration in the other channels making them hard to correct… you can using TE and blending with channels…
This is no with all defaults
Maxrgb
Luminance
powernorm
EUCLIDEAN
EDIT : Note no artifacts at the tip as scene in your image which are created by your extreme CA settings all maxed out…
Looking closer you have also cranked the chroma and saturation out of gamut and the setting you use in the CA module have created most of that blue tip…add that to the curve you use and I am not sure what you really expect to see?? Maybe explain your "its not working right " in a bit more detail…
2 Likes
@Newerth
First, the manual clearly states that there is no one proper setting for “preserve chrominance”.
Second, using normal settings is not “working around an issue”. Using settings that break the image in other ways could be considered creating one, though.
If that’s how you want to discuss failings in a module, good luck, and have fun
Well… I don’t know what to do now. I tried to crank all the values so that the issue would be more obvious. I also went with the CA module as high as I could so nobody could say I have to compensate that before filmic comes in. Now all of it is to much apparently so I’m told I’m doing it wrong again. How shall I proceed?
This is hilarious. You’ve quoted my comments that’s why I’m in these threads… Trying to set misunderstandings/misrepresentations straight.
Those concerned about the attempts at constructive criticisms of some of filmics behaviour should have a look at how aurelien, completely unprovoked, behaves towards other contributors.
For those looking for samples should have a look up thread. Darktable's Filmic RGB defaults render this desaturated - #8 by Newerth
There are also lots of side by sides in the Sigmoid thread see this post for instance. Note that the sigmoid RGB ratio looks very much off as well. New Sigmoid Scene to Display mapping - #545 by jandren
Most files will show it but portraits will reveal it more clearly. If people don’t see it from these samples and can’t recognize it in other files I don’t know what to say. It’s not as simple as desaturated on straightforwardly tinted. @jandren s synthetic samples probably reveal something for those that don’t see it in photos. I’m surprised so little difference can be seen in the synthetic midtones because in photos the effect appears to be very prominent there. It could be one of the many visual pitfalls I guess, the actual issue could lie elsewhere.
Looking at your image the default CA corrects it nicely but you used such a large radius you created a massive artifact…The authour rawfiner does a nice job working through the CA workflow. Using first raw CA then manual TCA on the lens and the the new CA module…normally you would need all those steps…as for filmic…you have few blown highlights …not too many but the norms exaggerate things in the extreme highlights at least that is what I experience. The new v6 filmic has a slider to tweak extreme luminance saturation so that can help…On your image testing I just used auto exposure and the default filmic settings…v6 now uses max RGB but you can use any…you can see the norms working if you pull back the exposure on your image…look at the original data and then you can see what it does…
1 Like
This is your image with the exposure dialed down to show the original data with only WB…note the blue channel…
Adding filmic starting with no…
Now see how the preservation modes attempt to restore that…depending on which one determines the final look…
Maxrgb
Luminance
Powernorm
You get the idea…
This problem can come in extreme highlights with clipped or near clipped channels as not all the information if there to do the best correction…although I believe just in the last day or so AP added an adjustment to the math and this should help to handle them better…
It was just in the example you used you had a few extreme settings and these will be compounded by filmic. You can for some edits you can add things after filmic so that you don’t have these sorts of interactions as well…
Well, is the look of the rest of the image the result you want?
If not, start by showing the issue in an image that is otherwise what you want. If it’s not visible there, even when knowing what to look for, is there an issue?
In addition, in this particular case, the manual states that there is no right setting for the chrominance preservation, so you have to pick the one best suited to your image. If that is “no”, so be it…
Everything can break if you push hard enough. That’s not a fault of the tool, though, if it’s used outside design parameters.
1 Like
Thank you for the elaboration. I’ll look forward to the v6 then. I wouldn’t to drop into this discussion again if there wasn’t for @aurelien’s comment. It’s just not as straightforward.
Have you used v6 yet…in response to much of the cries from those not liking flat images it now does not desaturate. I have played now with both modules sigmoid and filmic v6 and for the most part I can replicate quite easily the sigmoid look now. There are still the odd image where possibly each one out does the other but IMO there really is not anything massive there. And just the other day AP has tweaked the math for the norms when dealing with clipped channels I believe so it should be even better…
I agree the tone and exchanges in emails are often quite harsh but I guess others are free to take the DT code and run with it as has been done with RT and ART if there are features that they really really need…
I’m sorry – you just don’t get it. Please pay attention to this strip of text from a post above:
here’s one taken to extreme because I don’t have a better recent example
If you’re having a rough morning, please understand that I’m here to get help, not to be told I’m an idiot.
1 Like
I am trying to remember if you were on windows?? If so you can download Bill’s windows insider weekly build if you want to try it out…
Not yet. I’d rather wait for the official release. I just don’t really want to use a preproduction sample on daily basis. 
Thank you for the samples though.
NP I run multiple versions for testing…its nice to be able to test the features…most thing making the master are pretty stable but I hear you…
2 Likes
I think I do get it, I just do not agree with your methodology as shown in the post. where you have pushed almost everything you used to extreme, and for me unrealistic, values.
And if I thought you were an idiot, I wouldn’t bother replying. I’ve learned a long time ago that that’s just an efficient way of wasting my time (lots of it…)
Only very briefly. Opened a bunch of recent files of mine but the pastel look was immediately noticeable compared to RT despite colorfulness preset being used. Extreme colours like orange warning signs at road works turned toward red with the latter preset.
I’ve been emphasizing that my comments are not about “punch” or straight forward desaturation. For many years my photos were desaturated compared to the scene now I’m inching closer to realism from the flat side.
My short advice would be this …leave it on the defaults. Adjust white and black relative as needed and less often contrast. You can certainly try one of the other preservation modes at any time but you can usually go with powernorm or Euclidean, although maxRGB seems to be the new default for v6. Use the waveform histogram as your editing guide as it can tell you a lot esp about the highlights. If you need to balance channels a bit there use the tone eq in those regions and tweak by blending in that color channel…ie if blue skies are clipped try to bring it back that way and then balance the whole sky.
If you watch Boris edit you will notice that he does not mess with filmic hardly at all and does a large part of the job with the TE and rgb colorbalance instances…
Maybe throw your flower image over to the playraw forum and see what comes back…then compare it to your approach and see what differs…
Just some random thoughts… You can also try your images with the base curve or tone curves. The pipeline if more friendly to the basecurve now…what ever works…many images can be edited without filmic…
Now the extreme highlight saturation slider becomes a nice tool to tweak this. I suspect actually that the rgb colorbalance preset might actually not be the best for the new filmic and it might need toned down…also the preservation modes seem more consistent now but it could be me thinking that…the looks could vary widely before but I don’t notice such a wide range of results with the different modes now…I made a couple of color match icc profiles (jpg raw pair) and I use this with autoexposure as of late as I find it a good combination and filmic. I add lens correction and NR and that is about it…I tweak color and brightness with color calibration and use color balance where needed but my profiles do a really nice job with color so basically I just tweak the exposure if the auto is a bit high or low and I’ll do a local edit or two with the TE and that covers most of the images I process at this point in time unless I need to do some damage control…
Well, @Newerth did say he had pushed that image to extremes to show the issue…
On a second look, I’m not sure the “chromatic abberation” is not just a tiny bit of subject movement (looking at the hairs on the bud just below the top leaf, wind perhaps?). And then it’s not surprising the chromatic abberation module needs to be pushed to extremes…