Blender AgX in darktable (proof of concept)

[Patrick_Duquoc]

The versions posted here and compiled by our friends are quite good to me.

[Dave22152 (Dave Goldberg)]

It might have been better if he had made a separate video on AgX…

[JankoK (Janko Kernc)]

Linux Mint, and the appimage file found on this post:

Blender AgX in darktable (proof of concept)

So, a new build: gamut (negative) controls using the algorithm from Blender’s AgX; they are applied in the ‘base’ space, just like in Blender the ‘tuned’ primaries are available as presets blender-like|tuned and blender-like|tuned punchy. Please test them; if there are no objections, I’d like to make them the default, as they seem to match Blender AgX results more closely (see above) the default for preserve hue is now 40% for all blender-like presets, as that is what the generator script u…

[kofa (István Kovács)]

OK, I’ll have a look. Thanks for reporting.

[betazoid (Anna)]

@Pascal_Obry

there is also the possibility that Agx won’t be in 5.4 nor in later versions

???

[kofa (István Kovács)]

I haven’t even raised a pull request. Without that, it’s sort of hard.
Then, the team will provide feedback, and I, with their help and guidance, will have to make sure the code is up to standard. Don’t forget that I’m not a C developer and not familiar with the code base and image processing in general. Some of the code (the UI-related part) was generated using an LLM.

Plus, as good as the feedback here is from an enthusiastic group of users, I never asked the development team if they think such a module would benefit darktable (we already have two tone mappers to be used with the scene-referred workflow). A new module means more possibilities, but also more complexity, more maintenance. It needs commitment on my part, and a whole lot of trust from the team’s part. And time.

[kofa (István Kovács)]

OK. So based on the last polls, it seems most are OK with the current functionality and controls. From now on, we are in ‘feature freeze’ for the module, and I’m working on clean-ups only, to be able to raise that PR in a few days.

[PD1 (PD)]

@Pascal_Obry

Yet, that’s not nice to do that in a 5.2 presentation because there is also the possibility that Agx won’t be in 5.4 nor in later versions.

@kofa

Plus, as good the feedback here is from an enthusiastic group of users, I never asked the development team if they think such a module would benefit darktable (we already have two tone mappers to be used with the scene-referred workflow). A new module means more possibilities, but also more complexity, more maintenance. It needs commitment on my part,

I do hope the developers will account the user feedback of this discussion thread, and the utility of the AgX module as has emerged in the discussions, and not confine to the technical space only, like maintenance of code etc.

[kofa (István Kovács)]

I think they have demonstrated a positive attitude, e.g.

Blender AgX in darktable (proof of concept)

It’s a bit early indeed, but yes I had some thought about this since I’m following this thread. My thinking at the moment is that AgX should be the default, but I have not yet tested it so everything is still possible. If we go for AgX as default it should be done anyway after a full release cycle to get 6 months feedback on it’s usage.

[MStraeten (MartinSt)]

it’s just about being ready for production or not - there’s no feature roadmap, just due dates for feature freeze etc.
If an implementation is done and properly tested, it can make its way into a release, but there’s usually no intention to include green stuff just because of user’s request …
… unmaintainable code is quite green stuff

[nwinspeare (Nicolas Winspeare)]

Hello and sorry if there is any confusion. I think I say that I hope that agx will make it into the next stable version. My opinion (that seems to reflect all the enthusiasm in this discussion) is that it is a great module and would be an opportunity for darktable to move forward by combining the best sides of filmic and sigmoid.
If it stays as an experiment then apart from the disappointment, I’m sure viewers will get over it !
I also ask people to come and read this discussion for more information.
Hope I am not overstepping.

PS : The changes in the tone equaliser with the 2 sliders in the centre tab were brilliant and I hope that makes in through too.

I realise that Pascal has a difficult job and I understand the moderation in his comments. Thanks to all the developers and the entire team for your tireless efforts.

[europlatus]

@kofa

Plus, as good as the feedback here is from an enthusiastic group of users,

@Blender AgX in darktable (proof of concept)

If we go for AgX as default it should be done anyway after a full release cycle to get 6 months feedback on it’s usage.

This 6-months feedback on its usage is a very important point. A related point about user testing was raised in the other thread about the potential Darktable rebrand. The video linked to in that thread talks about the value of “new user testing”, which is basically putting the software in front of a new user and asking them to complete a task. The ease/difficulty in which they do it will give an idea of how well the UX has been designed.

So far, the feedback on the module has just been from the enthusiastic users active in this thread, and they tend to be technical and proficient users of Darktable. Getting more feedback from a wider panel of users and on a wider range of pictures will be invaluable, although I realize that it might cause additional pain for @kofa!

[Terry (Terry Pinfold)]

I for one will be disappointed if AgX doesn’t find its way into 5.3, but defer to the developers on the suitability of that happening. I feel talk of AgX being the default tone mapper is premature and what I like about Sigmoid as the default is that for many images it gives pleasing results straight out of the box with no input from the user. This makes it much easier for new users to embrace darktable.

AgX in my very limited tests felt to me as a first time user of AgX as a mature evolution of filmic. But I found myself longing for a “feature freeze”, inclusion in 5.3 and some documentation to explain the sliders and options to me. I feel @kofa has done a great job developing this module and the feedback from users is very positive.

[hatsnp (Nuno Paulino)]

I agree with Terry. In my opinion it doesn’t make sense to test AgX on new users at all as a way to make it friendlier whilst at the same time ending up risking making it less powerful.

New users have the perfect tool for their skill level which is Sigmoid. After, with what they learned with sigmoid, they will be more apt to tackle AgX and filmic (which is still the most challenging tone mapper to get a good image out of imo.

[MStraeten (MartinSt)]

to all reasoning about agx in 5.4 or not:
no one thinks about holding something back, but keep in mind:
Grass doesn’t grow faster if you pull it …

[kofa (István Kovács)]

I always said I’d raise the PR once the prototype is ready for that. The core team will approve it if and when it is ready for that.

It would be unfair to put pressure on developers (this time I’m including myself) to get it merged. No one (no user and no developer) wants an immature module, one that’s either functionally broken, or is impossible to properly maintain because of spaghetti code.

[Popanz (Uli)]

It’s ready when it is ready. As long as it comes to darktable, I’m happy.

[difrkaguilar (Franklin Aguilar Matos)]

If someone is impatient or in a hurry to use AgX, they can do it using the test versions that are present in this thread and that have been generously ported to Windows and Mac. The version of AgX that @kofa has created and polished little by little with the tips and comments of other users, is good, but it is not yet ready to be part of the main code of darktable, besides, there is Sigmoid which is excellent for almost all types of images to be developed, and there is still much to learn from Sigmoid and other tools.

[rvietor]

@difrkaguilar

If someone is impatient or in a hurry to use AgX, they can do it using the test versions that are present in this thread and that have been generously ported to Windows and Mac. The version of AgX that @kofa has created and polished little by little with the tips and comments of other users, is good, but it is not yet ready to be part of the main code of darktable,

If the current version of AgX is not ready for inclusion in darktable main, there is a risk that any edits done now are incompatible with darktable once AgX is included (for different reasons, not just related to AgX). So using the test versions has a risk, that users need to be aware of.

---

Also, continuously pushing for inclusion is not going to help. There is a process for including a new module which takes time, not the least because all the work is done in the spare time of the devs. So making the devs explain again and again that inclusion in “the next official version” is not possible, only wastes the time they do want to spend on dt…

[Macchiato17 (Lars Fiedler)]

I would like to summarize and give a suggestion: it is clear, that

• test versions have by nature the side effect that edits made in such may break in final versions
• there is no time-frame planned for integration into main, it may also turn out that AgX never gets incorporated there

So I would like to encourage all of those being very motivated in contributing here (which I really like) to focus on the items or issues left and keep out the obvious as mentioned above.

I (and I think we all) appreciate the efforts that @kofa and the dev team put into this wonderful software. So let’s please be sensitive to their limited time and energy resources by leaving out reheating already discussed items again and again.

I wish all of you a wonderful day & weekend