Hi Mica, thank you for your support
Basically, I open RT/ART, put the 3x Raw into the waiting queue and initiate the post prod, no change, no correction (-> as a consequence having a look to the screenshoots, you may notice my AWB 5D faced a “kind of” problem at the time of the shooting 
).
Default settings difference is a good investigating direction indeed
Hello ggbutcher,
Thank you for the idea, that you imply I have to investigate into the defaut settings/prefs…
This photo is sold so I cannot share the RAW but I upload another triplet that face the same challenge
200722_16h59-9214.CR2 (23.5 MB) 200722_16h59-9215.CR2 (25.2 MB) 200722_16h59-9216.CR2 (27.6 MB)
What I did in ART: opening the directory containing the photos in the file browser, directly putting the photos in the queue, process queue . Then Highlight recovery is OFF and you get the purple zones.
What you should do : create a profile dedicated to this work
- open one image and apply neutral (if you want avoid any processing)
- enable Highlight recovery/blend or color propagation.
- to recover completely the highlights you should set tone equalizer/white=-40
- select camera input profile “color management/auto-matched camera profile” unless your colors risk to be borked
- demosaicing rcd+vng4
- chromatic aberration correction/automatic
- close photo ( it creates a profile alongside the photo)
- in browser copy paste the profile on all photos and send to queue
- you can also save the profile in <yourconfigfolder/profiles> and use it in the queue: apply the following profile on export
- you can also set this profile as your default profile
Hello gaaned92,
Thank you for the time you have spent on the serie to achieve the best way to get rid of the purple color. I certainly will compare your procedure to mine then I can improve.
That being said, the question I have is to better understand why ART is generating purple with default settings and not RT.
I reproduce the process with Highlight reconstruction disabled (and tonal curve set to linear to increase the effect). See capture. I joigned the image params as well.
200722_16h59-9216.CR2.arp (10.2 KB) 200722_16h59-9216.CR2.pp3 (11.2 KB)
Wouldn’t it be in relationship to the fact that RT relies on dcraw while ART uses libraw?
Hi,
This is expected – see the explanation by @ggbutcher. If the highlights are clipped, you commonly get magenta highlights. To correct them, use highlight reconstruction.
And btw art doesn’t use libraw yet – it might in the future but right now it shares the decoding part with RT
Using ART, simply setting Highlight reconstruction to “Blend” removes the magenta in highligths.
Left: no reconstruction - Right: with highlight reconstruction in Blend mode
Hello all,
It looks like we’re not fully aligned on the topic but as the comparison of the result achieved between RT and ART in default mode doesn’t seems to be relevant to anyone, it leads me to the conclusion that this is not important.
Thank you anyway!
François
The type and the order of some processes in the pipeline are definitely different between RT and ART, so yes you can expect differences when you compare both.
But if you think we missed your point, you can write in French… Peut-être que nous nous comprendrons mieux ainsi 
Edit: rewrite of post
When sending a photo with NO profile to the queue:
-
RT applies the default profile as it is defined in the preferences.
-
ART also applies the default profile as it is defined in the preferences. select as default profile “auto-matched curve” or a profile built by you.
But it seems that, in both cases RT and ART, if you defined as default profile a dynamic profile, it is not selected in the queue and the neutral profile is used.
In RT you don’t get the purple zones as “clip OOG” is selected in neutral profile,
If you want to apply dynamic profile in ART you will have in browser to select images and apply “processing profile operations/reset to default”.
Can somebody verify above observations?
Actually, what that difference and this discussion did for me was to finally clarify my understanding of what highlight reconstruction does regarding the so-called ‘clipped channels’. I incorporated RT highlight reconstruction in my hack software, and even with patient explanation by @CarVac I didn’t quite get it until now… So, en francais, vive la différence!
(Disclaimer: my French is South Louisiana Cajun, extremely limited, and I had to do this sentence with the assistance of Google Translate… 
)
@fedelin could you upload your RT and ART default profiles as defined in preferences/image processing/default profile?
Thanks
Eh bien bonjour à tous et merci pour vos réponses. Tout d’abord une chose : si j’avais pris plus de temps dans les calculs d’exposition de mon bracketing, je n’aurais pas eu de clipping 
Merci à gaaned92 pour la pertinence du dernier post qui m’a inspiré, comme in fine je vais créer un HDR j’ai généré des tif 16 bits avec une courbe tonale linéaire et aucun autre changement (comme je fais d’habitude quand les poses sont correctement étagées ) Et là, pas de mauve non plus à l’assemblage des tif provenants de ART.
En revanche, le constat reproductible que je fais est le suivant:
Les 3 .CR2 bracketes sont copiés directement du navigateur à la file d’attente (pas de fichier profile donc). Les jpeg issus de ART rendent du mauve sur les zones cramées et pas ceux de RT.
L’idée est que peut-être cette observation peut apporter une amélioration, à tout le moins une réflexion pour ce fork qui est vraiment une réussite !
Cordialement,
François
J’ai bien compris.
MAIS, RT ou ART appliquent alors le profile par defaut defini par l’utilisateur sauf si le profile par defaut est “dynamic” auquel cas aucun profil n’est associe.
Or, Rawtherapee definit au premier demarrage un profile par défaut “auto matched curve ISO low” alors qu’ART définit un profile “dynamic”.
Donc en envoyant directement à la file d’attente si vous n’avez pas redéfini le default profile:
- Rawtherapee appliquera un profile et tout sera ok
- ART n’appliquera aucun profile (neutral) car le defaut est “dynamic” et il n’y aura pas de reconstruction des hautes lumieres. d’ou la couleur rose.
Voila l’explication de la différence.
Je vous propose donc pour resoudre le probleme dans ART, de choisir dans “preferences/traitement de l’image/profile detraitement par defaut/raw” le profile " auto-matched-curve".
@agriggio wouldn’t it be useful to apply default Dynamic profile when sending directly raw from file browser to queue, as it is for other profiles?
Definitely – the current behaviour is not intentional, it’s a bug… I’ll work on it
1 Like
Cela règle clairement le problème, avec ce profile ART fournit un résultat identique à celui de RT. More than ever “when it works, we learn nothing”! Thank you all for the time you spent to investigate.
1 Like
should be done now
1 Like
@fedelin si jamais votre systeme est windows 10 vous trouverez ici les builds mis a jour : ART-W64NightlyBuilds/ – Keybase.pub
Choisissez generic , utilisez un dezippeur (7-zip par ex) extrayez le répertoire ou vous voulez, executer ART.exe.