Interesting points @Steven_Adler1 .
I just asked a “friend” about royalty free, assuming the concern would be a contributor subsequently asking for payment for work/code done.
Github terms of service state things like whenever you make a contribution to a repository containing notice of a license, you license your contribution under those same terms (apparently!).
Re. copyright, the contributor does retain the copyright to the specific code they wrote. However, by submitting the code to a licensed open-source project via GitHub, they have granted the project owner an irrevocable license to use, modify, and distribute that code according to the project’s license. They cannot retroactively revoke this license just because the project owner refuses to pay them (apparently).
If the contributor was given to understand they’d be rewarded for their contribution then of course it’s different.
I’ve wondered previously how much security checking goes on but life is not perfect, I’m not prepared to do checking (even if I knew how), so I just use the appimages and hope they’re not doing anything naughty!
If you mean the source code: that’s how open source has always worked (see @RawConvert’s note about knowingly submitting code to a repo vs royalties).
Security checks are done as part of the normal code review (people can’t just add random stuff to darktable, only the maintainers can do that). That’s what PR’s (‘pull requests’) are for.
Neither of these has ANYTHING to do with AI, though.
If you mean the model development: big companies just harvest images from the cloud. Open-source developers often use their own and others’ generously contributed photos (e.g. Natural Image Noise Dataset - Wikimedia Commons), much like how https://raw.pixls.us/ or how Lensfun receive such images.
a feature flag is present, thanks! (@anry that you have to enable in the settings it not really a feature flag in the traditional sense)
I understand that is being worked on, either by the model cards mentioned or any other way. Thanks!! (@rvietor a link to more information is not really a nightmare as long as the page is under control of darktable, so the darktable docs or webpage will work fine)
Then number 2. Is it possible / desirable that the darktable website include the ‘ai-options-not-included’ builds?
@Steven_Adler1 I think that the discussion about security checks etc is interesting but what to leave it out of this thread.
This answer indicates that submissions to GitHub have license terms. Good.
@kofa indicates in a reply that security checks are done as part of the normal code review.
I worked for a big company and we ALWAYS checked royalty free licenses and security code reviews because we were sued and cyber-attacked often.
@kofa I raised these issues in a post to say that ethics reviews of AI training data are nothing different than cyber-security code reviews and royalty free licenses. It’s a compliance condition that might seem new and strange until you do it regularly and then it becomes normal.
/*
This file is part of darktable,
Copyright (C) 2025 darktable developers.
darktable is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
darktable is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with darktable. If not, see <http://www.gnu.org/licenses/>.
*/
so a contributor is aware that he provides his contribution under a gpl license.
if not then i doubt he’s able to contribute something acceptable
A big company has a different business model. They try to avoid “infective” licenced code because that might destroy their plan of making money from intellectual property.
Maybe read the GPL to understag the differences …
This thread almost makes me feel like the devs had it all planned out before they merged the PR and I made all those long posts for nothing …smh
But it’s good to know that my concerns weren’t for nothing and the implementation takes care of them. The usage of certain models may always be a topic of debate but then again people will use whatever they want to use once they install it on their machines and I get the “freedom” part of the argument from the devs and although I want people to prove that great models can be built with the help of the community instead of stealing from them, the hardwork required and the sheer scale of things makes me wonder if that is part of the problem why it is easier to accept other things and hence leading to the same loop of arguments.
Again I have massive respect for the devs and It is almost impossible to do things according to one side so a balance between the two is appreciated.
I know a few people who love darktable but at the mention of Ai they almost immediately responded with “that makes me sad” and started to mention they might have to rethink if they will keep using it or not. The “optional” part everywhere and simply how modular darktable is, makes them more likely to love using it so I can confirm that much.
The other side is “do i need to use some command line cli tools/scripts? is it in the GUI?” and when I mentioned “you download your models either from the repo or from the GUI” makes it a lot more appealing.
I guess You have found a nice balance where everyone may not be the happiest but most will be equally happy.
(again most haha)
“no ai” vs “oh why won’t it ship with 5.6?” will not end for the time being I guess.
They are sad maybe because they don’t know what AI has been added into Darktable. We have only:
AI to create a mask - that is create a drawn path
AI to remove noise
AI to upscale image
There is nothing (and there will never be as long as I’m maintainer) like removing people, suppressing background, changing sky or adding a cat… into the image.
As already said many many times, we can’t talk of just AI to express all the different usages.
I’m a photographer and certainly don’t want Darktable to change drastically my picture because I’m lazy taking real picture. But I can understand the confusion as there is a huge amount of messages saying just stupid/wrong things.
And please don’t make this thread as useless as the other on the same subject.
I think a large part of it is simply people not understanding that “AI” can be several different things, many of which have little, if anything, to do with generative AI like ChatGPT and friends.
Mostly from people trying to jump on the AI hype train I’m guessing. There’s been machine learning driven models in software for a long time. Now suddenly if you slap an AI label on the can in comes the VC dollars, euros, pounds, yuan or whatever. It’s really clouded the software world and made everything more polarized.
People have a visceral reaction to what is now labeled as AI one way or the other, myself included sometimes, mostly from the marketing saturation and the MBA fan fiction the Big Tech guys write about laying everyone off because of it.
I mostly agree. I have worked on recognising human emotions by software using machine learning. But this is already more than 20 years ago!!
Though, the last 4/5 years something has changed dramatically. And that is not to be underestimated. We have had a great increase in compute capabilities and now have the ability in the machine learning applications to do it more or less ‘real time’.
Our application took 5 minutes or more compute. But now the emotion recognition can be done in seconds. And image generation etc, is now a couple minutes.
Additionally, 20 years ago it was impossible to crunch the data to train the models we are now using.
So, AI is a marketing term, greatly over and misused. But a big leap is made in machine learning algorithms compared with 20 years ago. Mostly by the huge increase in compute power.
