You are so right…Alberto is trying to see if he can also convert this one…
JuanPablo’s github site has a treasure trove collection of tonemapping work…
3 Likes
I haven’t added the matrix tweaks yet, but here is a half-done port of the gamut compressor from @jedsmith:
7 Likes
The thing I like most about AgX is the way it handles the highlight rolloff from clipped areas. In an actual digital photo you often get ringing artifacts around a blown out light source, and AgX seems to minimize/eliminate those in am extremely pleasing manner. I don’t see the same kind of artifacts in these 3d rendered images. So while they provide a nice test case, almost nobody will have images like that in real life & testing against real images is still important.
4 Likes
I think that gamut compressor should get its own module, coming just before output color profile. It’s 6 sliders (compression sliders for the CMY complementaries, plus threshold sliders for RGB to specify the starting point of the compression). Maybe it’s time to open another topic. 
5 Likes
Yes the gamut handcuffs in filmic v7 seem to throw a lot of people off and makes troubleshooting and analyzing things more difficult than it needs to be… let’s not make that mistake again!
1 Like
Looking very nice right here!
1 Like
Been there, done that, just today. Editing a … load of shots I’ve taken around mid-day w. strong contrasts, after some experiments with agx I went straight back to Sigmoid. Nothing against agx at all but I just need more practise with that newer module and rather wait for a final version before I use it for those time consuming jobs.
BTW … having shot slide film for the better part of my life (nailing exposure and frame in-cam) I only use tonemappers if needed, for shots without those strong contrasts I prefer to stay with the usual suspects. YMMV.
1 Like
You’re probably not looking for feedback, but I did notice some harsh transitions and abrupt colour changes on the picture of the children:
But this could maybe be mitigated with some desaturation, as the desaturated AgX version that Billal posted earlier was the best of the bunch. And of course, I realize that this just a half-done implementation of what you are planning.
The sigmoid edits above are similar, but slightly less saturated. Again, synthetic images 
@s7habo am I guessing which one is which??
This is looking beautiful for a half done job.
I was just following the clipping indicator, what confused me that I was observing the waveform when comparing the two and they are literally the same but somehow the image looks more saturated in Sigmoid and the indicator is not signaling clipping but thankfully @kofa is making great effort along @flannelhead to provide with amazing solutions.
Even when editing real life examples it does “break” rather quickly when increasing contrast for example.
Oh, that’s only without and with Agx. Nothing special 
I forgot to mention it.
When kofa gets done tinkering, I’ll be happy to compare all three tonemappers. There is no lack of photos
Yes, but you can use the rgb primaries to occasionally adjust the purity of primary colors. Then it actually works very well.
Here is the example with high contrast:
And with reduction of purity of blue primary:
2 Likes
I was also comparing exposure sweeps to compare between the two and I think Agx hold really good at high levels. 0 EV to 6 EV
Only the lightsaber picture is 3d rendered, the other ones are made from raws from real cinema cameras, as far as I know. I don’t know the full process how the EXRs were made, but I think it’s just demosaic, color space conversion and white balancing.
By the way, one slight quirk that darktable has with these EXRs: since they are encoded in the Rec.709 primaries, they will contain some negative values. However, at least in the darkroom darktable clips the negatives off very early in the pipeline when scaling down the image. You’ll notice that the image changes when you zoom in to 100 % in the darkroom, as the clipping is not applied then. I’m not sure if the same problem exists in export. Just something to be mindful about @kofa and @Billal
3 Likes
@hannoschwalm I think this is the reason for the differences (HQ/zoomed in vs not) I mentioned on IRC the other day. Thanks, Sakari!
Yes, that might be. (But clipping to >= 0 happens elsewhere too)