Ansel “Filmique V8(Agx low bleach)” vs “Darktable AGX” - Sunset
Source : PIXLS.US Sunset on ferry ([PlayRaw] Sunset on ferry)
Image : https://d2x313g9lpht1q.cloudfront.net/optimized/2X/8/8bf07bf404bacc8a2771f01025212e9eb1ad9c5a_2_690x460.jpg
An image which clearly highlights the difference in behavior of the two tone mappers.
AgX is best, the “pink” color around the sun of Filmic is clearly not something you’ll see on a sunset.
This is a rather unfair comparison, as both modules include a variety of setting that can make one image look like the other.
You could compare default settings / starting point, but that’s also comparing apples to oranges a bit.
Are you sure, this is a proper comparison? It looks like on the ANsel edit the highlight reconstruction is not set properly. And the question is, what do you compare? Standard settings? Or have you tried to achieve a similar result? I prefer darktable over Ansel. But I think this comparison doesn’t do Ansel right.
Is Filmique V8 a new version of filmic rgb? You don‘t use filmic rgb for sunsets.
Maybe not a fair comparison but I was interested to find out that Filmic V8 is available. Hopefully it can be added to DT?
The example is an implementation of AgX within the filmic module of Ansel
Filmic has always been the weakest at intense highlights. Even if it can be tuned, it’s not intuitive and still doesn’t give ideal results.
Why add it to dt? Why the FOMO? Is it better?
I downloaded Ansel and could achieve a pleasing result with it, but I have to say it was quite difficult. This was more caused by the fact that Ansel hasn’t moved on with highlight reconstruction. To get to a for me pleasing result, I had to manipulate the white point. Still I wanted to get a bit different highlight roll off, which I wasn’t able to achieve. But I’m a dt user not an Ansel user, so maybe I’m the problem, not the editor.
Further observations:
I second the opinion that this comparison (or at least these screenshots) demonstrate nothing.
In principle you could swap the results by changing the parameters in the 2 modules. As things stand, we can’t know why you chose those settings.
Plus, tone mappers do not work in a vacuum, they interact with the rest of the stack.
How can it be? We’ve done so many CPU/GPU optimizations those last years.
9 posts were split to a new topic: darktable nightly slow?
Exactly the point to me 
The guided laplacian highlights would need quite a big radius here. That image btw has been in the highlights-reconstruction-testsuite for long. You can also spot the incorrect results at the white wall to the left.
The highlights mode in filmic has not/never been intended to handle such color casts introduced very early in the pipe. BTW - it’s not at all the tonemapper’s job i would say (as it’s by far too late in the pipe).
I don’t dislike filmic, but all the other points match my observations. With its improved ui (mostly) and updated colour science on familiar modules (color eq, color balance rgb, etc…) I wanted to really like ansel, but missing some dt favourites is a hard sell. Highlight reconstruction is at the top of the list.
My understanding of v8 in ansel is that instead of adjusting latitude with one control it can now be done with two, one for shadows and another for highlights, to mimic the toe and shoulder controls of agx.
AMD or Nvidia?
Not really, it highlights the difference in your use of the two tone mappers.
It is an instance of a hasty generalisation
It highlights that you have to prepare for corrected data asap in the pipe. Highlights, capture sharpen …
I am not a statistician (nor do I play one on TV). However, I used to work next door to a research statistics group. The lengths they went to, to minimise the effects of confounders never failed to impress me.
Not fear of missing out. Film was AP’s baby and if he has created a V8 I would like to see it included in DT where we have the choice of which version to use. I rarely use filmic but I like options.
