No, this is not an “AI Bad” post, on the contrary, I’m looking forward to more AI.
It’s really cool that sam2 object segmentation is available, and I’m sure that it will be improved further because it’s slow as molasses on my relatively good computer.
But what does integrate more AI mean? Does this mean new means of selecting something and only apply effect to that area will be a thing? If so, looking forward to it.
Most do not want AI cause it is currently being misused more often than not.
Who is this “most” you’re referring to?
You’re not changing my mind, so don’t bother.
There’s nothing wrong with AI, first of all. It’s a tool. So tell me, what is wrong with AI being integrated into kdenlive. It’s already here in the form of object segmentation mask making through meta’s SAM2 library.
For anyone stumbling into this thread wondering what I’m talking about… the roadmap says midterm that the devs have approved the idea of utilizing AI into kdenlive and it’s going to happen sometime down the line. Most likely not generative AI, but similar to krita’s planned neural network they want to integrate that doesn’t generate anything but instead automates extremely repetitive tasks for you.
Screenshot if you can’t access the link.
Like how environmentally unsound it is, which goes against KDE’s Eco commitment. Then there’s the copyright infringement. Then there’s the massive amounts of CSAM that is being generated with it.
The only thing it seems to be good at is generating AI slop images, and made up stories to post to AITAH on Reddit for rage bait. The code it generates still needs to be manually reviewed. The music it generates is awful. The “art” it generates sucks and the people who think they’re “artists” because they now how to prompt a computer are insufferable.
I have no problem with AI being a separate stand alone package that people can install themselves after the fact. But if you start pushing that crap into my desktop by default, I might as well go back to Windows and subject myself to crap I didn’t want or ask for that is of absolutely no use to me as an end user.
You are mistaken that I am talking about Generative AI like Stable Diffusion or Midjourney. I fully agree with you that generative AI is terrible, but not all AI is midjourney. It would be more appropriate to perhaps use the term “neural network” because it’s not making a single thing, not a single new pixel is generated.
Krita for instance, is currently in the process of of doing a line art cleanup functionality. It doesn’t generate anything new, you feed in a sketch and it just applies the ink to the exact pixel it was fed in. If you give it a messy sketch, you’ll get a messy inking. That’s the kind of thing the KDE devs want to do: just a simple algorithm to help the extremely repetitive tasks and making your life easier.
I’ve already mentioned the SAM2, which is yes, AI. Made by meta (facebook). It doesn’t generate anything new at all, all it does is just track something you apply a boundary to and automatically apply a mask to it eliminating the extraordinarily tedious way of doing it manually.
I do wish the AI is evil mentality would fade, because it ruins the ability to even talk about it for a positive use. I don’t blame you for mistaking this for wanting generative AI, because the AI advocates most commonly use it to infringe on copyright so they can make “new” stuff.
@northlean AI in general needs to be slow rolled, very very very slow rolled. It is bad fore the planet, it is bad for our freedoms.
Well, as far as freedom is concerned, the SAM2 library is a combination of apache 2.0 and BSD 3 clause (Also referred to as “new BSD”) licenses. Even stallman himself, who is known for being unmovable on software issues, wouldn’t have a problem.
You’ve spent quite a long time telling me that AI is bad for the environment which someone else gave a better response, that it’s bad for freedom. How though?
A better question would be, why is KDE trusting meta of all companies though? I don’t know if I should (sorry about this) ping a dev such as @meven or someone else. If meven knows a kdenlive dev, please tell me. Thank you.
Indeed AI is like any technology isn’t bad by itself only its use.
Just like cars are fine, but bank robbery involving cars isn’t. Still like with cars simply using them has some consequences (gaz emissions, accidents…).
In case you missed a few points:
LLM straight up make you loose some skills and less good at judgement, so say a Microsoft study:
It can introduce discrimination, when the training material is biased, it can be built-in racist, or ignore cultures and languages.
Or infringe copyright, when an AI spits out book or code without the consent of its authors.
It can be used to threaten democracy, make up information with deepfakes…
It does people’s time or change the perception of reality with AI Slop.
And LLM do hallucinate by design and always will, you can’t trust what it outputs.
And did I mention how much electricity this consumes, increasing gaz emission in the process. Because it is based on brute force. NPU can improve 10% but when data crunched increase 10x that is still 10x more power consumed…
An example where it might make sense:
https://lwn.net/Articles/1026558/
And cancer research…
I.e specified areas where experts DO review what LLM spits out and within a restricted domain.
That statement makes little sense.
“KDE” isn’t trusting any company. KDE is a community made of its contributors mainly (https://timeline.kde.org/ Repositioning the KDE Brand - KDE.news).
In case you missed it, the community is environment conscious (Fin de Windows 10) and a FOSS advocate (https://manifesto.kde.org/).
Few LLM are really FOSS with training data available. Only GAMAM and startups do master general ones or have the resources to run them. Neither is it eco-friendly.
And it is expensive, ultimately for the users, making them dependent of those services.
KDE software has always been about empowering users make better use of their own computing.
Still usage choice should rest on users, as long as we have contributors willing to implement it.
Digikam just added usage of deep learning digiKam - digiKam 8.7.0 is released
We have connector to Microsoft onedrive and Google drive (until Google revoked it) for instance.
@northlean you can give it a rest cause you’re not changing my mind PERIOD. I have already muted and blocked you.
@TheTechieLife Sorry to intrude, but you seem like the person to ask: Is northlean’s way of thinking getting common around here? I’ve been using KDE for some years now, but I’m new to this forum and I don’t really know the community. I chose KDE because I thought it was environmentally conscious and pretty much aligned with my ideas, but if we’re starting to pivot to AI, I’ll need to start looking for other communities to join.
And yes, I’ve browsed the forum and I think we’re alright, but nothing beats asking an actual human for their informed opinion, so I’d appreciate it very much if you gave me your take on the matter.
Hello. While I agree with @meven, I think that the future of a desktop will necessarily be linked to an AI (a small AI without any skills other than language).
In the future, I see Plasma (and other) as a super MCP server:
I can imagine the enormous amount of work, so I don’t ask for anything; I simply dream of a future 
" in all applications"
so, the “mcp” serveur is Qt before plasma ?
“IT” better not be referring to me.
Please keep on topic. If you want to discuss if AI is bad or not or anything else not related to the topic of this thread, please create your own instead.
You are absolutely right, I apologize.
Can somebody please remove the kdenlive tag from this thread (and probably close it too).
It has nothing to do with kdenlive, and “AI” is a meaningless buzzword.
If people want to talk about specific technologies and tools that may be of interest or value to kdenlive users or video editing, that would be very welcome - but your dreams and or nightmares for “AI” (as if that was something more than the Titanium Enterprise Pro Plus badge du jour that marketing people are sure will increase sales), belong in the 1950’s - or if you just can’t stop yourself from larping that golden age of hysteria - maybe on fb or utube, or where all good pipe dreams and conspiracy theories are sold. There are plenty of other places for that sort of thing.
Please let’s keep this one for things about using kdenlive and actually editing video.
Since the OP has specifically asked about Kdenlive’s roadmap and has made it clear that they are referring to AI in the actual sense of the word (so not necessarily LLMs, but actual non-garbage-outputting AI technology), this post makes sense.
It just keeps getting responses from non-Kdenlive people and anti-LLM people for some reason.
Presumably OP’s thread would have gotten better responses if it was outlined something like this from the beginning:
Title:
Kdenlive plans to add more AI, what exactly?
Body:
It’s really cool that Kdenlive’s current SAM2 object segmentation AI tech is available, and I’m sure that it will be improved further because it’s slow as molasses on my relatively good computer.
The Kdenlive roadmap says midterm that the devs have approved the idea of utilizing AI into Kdenlive and it’s going to happen sometime down the line. Presumably not LLM AI.
What other AI technologies are the Kdenlive devs planning to integrate next?
(PS: The currently implemented AI technologies in Kdenlive, SAM2 and Whisper, are not LLM-based.)
So just wait until some Kdenlive dev actually replies.
When there is something specifically planned, we’ll specifically add it to the roadmap.
It just keeps getting responses from non-Kdenlive people and anti-LLM people for some reason.
Because if “Kdenlive people” had something to add, we’d have added it to the roadmap.
If this thread had been suggesting something specific we should look at, or actually asking for help with something kdenlive does or should do, there might have been something to discuss, but there isn’t, and it didn’t develop into that. So now it’s just a growing black hole attracting AI trolls and if not treated as such, will just continue to accrete more of them if we allow it to.
We tried “if we ignore it, maybe it will go away”, but evidently that isn’t succeeding.
