Note by moderator -
This thread is moved from the initial topic - Introducing a New Project: Fast Line Art - #56
Poeple discuss the merits and demerits of this feature in this thread.
@Minh is not the thread creator, it became the first thread due to the move.
Sad that making art is considered tedious by many here. I remember when it used to be fun 
I’d rather the efforts be directed towards having compacts UI for mobile, or improving performance, or like maybe just marketing 
The functionality is sponsored by Intel, then it’s this or nothing 
It’s not like money from donation has been spent for this functionnality.
If an AI function is implemented to improve the fill tool (like manage perfectly the gaps and short angle corners) it will be a good thing for many users.
But for some other that fill everything manually with brushes, they may consider that filling by hand was fun in the old times… 
If you read the topic, you can also see that for some users (disabled users for example) this can be an interesting tool.
And for users without a pen tablet too ; sketch paper, take a picture with your phone, ask Krita to trace your sketch.
Artists used to do it by hands will probably continue to do it by hand, to keep their own drawing style.
Grum999
I love to draw, but unfortunately when you have a lot of commissions time is a valuable resource and the faster you do things the faster you can take on more jobs.
This is not as helpful as you may think. This is essentially a replacement for doing line art, not a tool. Artists do not need this.
The process of art is not tedious - nor is line art.
“if the tool eliminates your choices or replaces them then it’s not a tool and that’s the fundamental problem with AI.
On each degree of agency you take away from the artist, the less it speaks about you.”
Regardless if it’s made with volunteering data, this replaces the whole process of doing line art.
Furthermore, Intel has been known to collaborate with / sponsor unethical (and illegal) projects such as Stable Diffusion, and we all know gAI uses stolen data by “copying” and working against the artists it has stolen from.
Another example of a similar piece of tech is Copainter, as @ramskulls mentioned. Copainter does indeed use Generative AI, and is trained on datasets with stolen data. Malicious people have taken sketches from artists and put them into Copainter; that is also a very likely possibility with this.
What about you improve robusts tools that will actually outlast this AI hype?
Like the 3D reference tool from CSP?
If you dont like doing line art, then don’t do it. But don’t make a piece of tech that takes away from doing art.
Also there are many other issues artists have asked for that have not been implemented yet, and this - we haven’t asked for. You also have previously mentioned you are against Generative AI - while this project is extremely similar to gAI.
““If you think “this is how AI should be used” have missed the point of genAI or similar ML. The whole point of it is to automate art, to make people dependent on it, to use it as part of the process so you stagnate.””
We don’t need automation for art.
I have been using Krita since 2019, and I love it. I believe you guys are not intentionally willing to do harm, but this project does more harm than good.
Please reconsider, keep this kind of AI away from your software. The fundamental way an AI-assisted filling tool is different from one who does the lineart is that the former can be done masterfully by a three year old. Some artists might just stop putting effort into learning how to draw lineart effectively, it’s a skill that needs to be trained to be able to do it masterfully and quickly. You won’t become Kim Jung Gi, Eliza Ivanova or Even Amundsen if you let AI do the lineart for you. Not to mention that this would make life so much easier for scammers who steal rough sketches to sell as their own work. Please, think long and hard before you do this, you’re about to pick a side here.
I’ll probably never use it, but I don’t even consider myself as an artist 
After, if some artists find it useful for specifics needs, why not?
If you don’t to use it, just don’t use it.
I already read a lot of topic here about users complaining about the stabilization tool that is not like this, not like that, not a perfect clone of some other professional software…
So if tomorrow someone implement a better stabilizer, something really amazing, but this stabilizer use some AI algorithm, is it a problem? 
Is it the tool itself or the way the tool is implemented that is a problem?
Doing digital art is already an automated process: stabilization, filling tool, line/curves/circles, assistants, selection, alpha channel, brushes engines, layers, blending modes, undo/redo, filters, …
I’ve already got remark few years ago that “didital painting” is not real art… “Real artists use paper, pencil, real brushes…”
This was a point of view similar to the point of view I’m currently reading here…
You didn’t asked for it.
Some users here seems to find this useful (not me, but I don’t care, it’s just something that I won’t use)
Some people (the sponsor) pay to get this functionality implemented, it’s not made from money get from users donations.
Find a sponsor with enough money to implemented a 3D reference tool, it will be implemented.
Also on my side I don’t see any difference between a 3D reference tool and line art tool: reading arguments, it’s pretty similar: being able to think the perspective and pose for a human character is something that needs to put effort during years to learn how to draw perfectly human body whatever the perspective and position.
Using a 3D model reference allows any dumb guy (like me, I’ve already tdid it ) to quickly get the pose and draw without putting much effort to learn how to manage a human body in space.
This is opened to debate… I don’t agree with this: on my side, I don’t use filling tool [with gap stuff] because I fill manually with brushes.
Does I consider that’s how a real artist should do?
No.
Because it’s my pleasure to fill manually.
And if people can spend time to something else than filling manually it’s a good thing: I’ve more pleasure to paint areas than drawing the lineart…
Then, if people can spend time to something else than doing the line art, where is the problem?
Artists that takes pleasure to draw their lineart themselves just won’t use the tool, like artists that don’t need filling gap tool won’t use it 
I’m not sure to understand but it looks like then there’s good way and bad way to be an “artist”…
It will be easier and faster for a scammer to just ask ChatGPT or any other similar product to generate an artwork from a prompt or any model rather than trying to steal an original sketch and produce a lineart from it…
Grum999
You ignored all my other points. Don’t try to justify tech aiming to replace the process of line art, or art in general.
If you think Digital Art is automation, or even coming close to AI, then you are completely mistaken. Using this “Fast Line” tech does NOT make you an artist. You made the sketch, but not the line art. That essentially strips away from your identity as an artist. You still need to draw, know how the layers work, know what colors to use etc… In the process of digital art, nothing is automated.
Stabilization, line/curves/circles. selections, brush engines, layer mode don’t automate the process for you. Brush engines are like simulation of real life brush strokes - not automation. Filling tool & selection tool uses the edges of lines and different colored pixels to detect edges - not automation, because you can also create an edge in real life using tape or a medium, and then spread paint into it.
Blending modes just alter the way colors are presented - not automation in any way
If it is a replacement, yes it is a problem; regardless if it is ML or not.
We are not asking for a perfect clone, we are asking for a similar feature.
You clearly didnt undertsand that comment.
If you use that piece of software, you lose the ability to do that specific skill yourself.
Also, what stops them from not making “auto-shading” a thing if they decide to continue with this. We didn’t ask for it, and we also don’t need replacements for art.
Do you know what ML is?
If you don’t plan on using such software, then there is no reason for you to defend it.
Back to the other points. You completely disregarded the fact that Intel collaborates with Stable diffusion.
How does this relate with the 3D reference tool? You still need to draw over the model, and have the knowledge of fundamentals. For line art, this does ALL OF THE JOB.
This is so depressing. I’ve been recommending Krita to everyone wanting to jump ship from Photoshop, but I won’t be doing that anymore.
And this is such a slippery slope. How do I know that down the line you aren’t going to start pulling the kind dishonest and exploitative stuff that Adobe does? I guess it’s alright because it’s “ethical”, even though it’s gonna absolutely wreck a lot of newer artists by automating an integral part of the art progress. I guess any excuse is ok to justify getting the big bucks from Intel.
I idealized Krita too much just because it’s free and open source.
And back in 2022, they said they were against gAI. I find that this project is extremely similar to gAI, and Intel is not on the artist side.
AI is a tool. It like a knife can be done for good and evil. And if krita does go rogue, someone else can fork it and remove it. This has happened in the past like XOrg vs XFree86 and XOrg won in the end. Same thing with GNU Emacs vs XEmacs and NetBSD vs OpenBSD, only with that last one both projects are still operating.
How many times do people have to say “AI is not a tool” for you to understand? gAI is not a tool, nor is this, as it is extremely similar to gAI. Also,
Choosing to go forward with this project can sound hypocritical.
Lets say krita does introduce gAI, someone can fork it and remove it like I said. I just don’t see a need to use rally up the torches and pitchforks. OpenOffice was the defacto libre office suite until it got bought by oracle. Everyone was not happy and forked it, nobody really uses OpenOffice anymore while libreoffice is thriving. That’s all I’m trying to say.
And how much good has been done so far for people who truly, genuinely care about the craft? This is so depressing, I thought I was safe with Krita.
I get your point, but that wasn’t my main one.
I said it’s hypocritical on their part, and the artists supporting the development fund would feel betrayed. The possibility of retracting their donations would be high.
Then what do you propose be done about it? Fork krita? Switch to something completely different like GIMP?
…Not implement this kind of tech in the first place?
If they are pro-artist, they wouldn’t implement it.
As I said earlier; I don’t think the Krita dev team wants to do any harm. That’s why we have to tell them the problems this project can cause.
If I were to paint every link in a chainmain and every pore in the skin by hand I’d die from hunger. My family would die from hunger.
Unless you’re one of the very few gallery artists, the ones people throw money at to execute every pattern and pore, you cannot afford to sink mind-boggling chunks of time into details “for the sake of art”.
Such is the life of working artists commissioned or in studios, expected to fulfill deadlines and keep their motivation high while fending off all the traditional obstacles to making a living out of their skill besides new threats like the malicious scraping of their portfolios (the very way we find work!!) done with the sole objective of replacing them.
I’ll not place the burden of fixing our predatory society on the backs of artists who can least bear it, the ones without safety nets to fall back into, for whom missing a deadline and payment mean they’ll go without meds, food, fail to pay for the accommodation and be pushed into further debt.
I share your contempt for generative images and the ones who did this to us. You’re clearly here out of concern, not trolling. I don’t give any corporations a pass either, if anything I judge them harsher than people because despite what laws may say people they are not.
Where we disagree is what Krita dev team is, how liable for entities’ actions they are, and how much trust can be placed on them. They’re not a corporation, are they? It’s ragtag group of people developing an open source software without the expectation of being handsomely compensated for it, pretty much the opposite of what corps are: entities created solely to make money, their very existence a pressure to do it no matter what.
And from what I see on these forums this team doesn’t take kindly to generative bull. It’s not sympathetic PR noise that matters, it’s the actions, and they’ve not allowed ai in this space, or even support for such tools here. They could refuse to offer official support but turn a blind eye to it being done by the community, but they didn’t, they chose to confront it. Contrast with what Artstation did, making room for them.
It’s also a small team of a niche software. Large, wealthy and well-established corporations like Intel can afford to make grossly unethical moves and still exist later, a small project? I don’t think so.
Even if they were engaging in bad-faith in this project I suspect they’re equally aware of this. Their userbase is the pressure that’ll keep further developments in check.
And I don’t know about you, I’m all for taking the money from large evil corps to do good and even subvert their lack of standards.
Intel, Adobe, Open AI, etc etc all used the labor of people in their datasets without consent. This project is the exact opposite, using only licensed material with explicit consent from their creators.
It has the potential to be a role model of how generative tools can be made ethically. The field needs it. We all know there’s no way back, but there’s legislation and a way forward in which training datasets consist exclusively of licensed and public domain material, in which they are tools, aids to create art, not replacements.
Ethical projects such as this have to be created in order to show an alternative use for this tech that is not out to hurt working humans. Otherwise all we have left are bad-faith grifters arguments that there’s no other way possible to train and use these models, but-but-but- the horses already bolted how do you expect to contact everyone used in the dataset? and a truckload of other gullible or dishonest points.
If this works then it’s the perfect counterpoint of “No, you don’t need to steal to develop this tech”, “No, these can be designed to be only tools”, “Yes, it’s possible to design them so they aid in the work without taking the creative decisions away from artists”.
I don’t think this tool will create perfect lineart, no offense meant to the dev team. It’ll stick very closely to the skecth (good!), meaning the sketch has to be good. It has to have energy or the output will look utterly lifeless. Even then it’ll probably create passable but not good good results, and certainly not true-to-one’s-style line weight and flow. So it’s an aid, not a replacement. The output will need to be worked over by the artist if it’s to have any character.
I agree there’s much to be done for other tools, basic functionality even. Unfortunately that’s not what’s being funded in this project. This lack of funding can’t be brushed aside like a minor detail when it’s the constraint determining what is or not being actively worked on.
I don’t think it is gAI… As it looks at your sketch lines and decides based on what is drawn. So it not generative.
I can see your concerns about learning lineart. But what if you just want lines to practice colouring? Sketch, lineart, done start colouring.
And I don’t think the lines will be precise enough to be able to satisfy the experienced artist.
I find your concerns reasonable, but than again I am wondering, what you are worried about.
Theft with gAI? It is already here.
People not learning art, because there is gAI? It is already here. There is even a generative AI plugin for Krita!
And now you are writing a wall of text, because the Devs are paid to develop it from a sponsor? I find that unreasonable.
Im sure you know there are brushes for that. You still have to use the brush, and you do it yourself. This doesn’t. This does it instead of you.
The problem arising here is that what if there were someone to provide data without the consent of the original artist?
Creatives know that at this point, we really don’t need gAI.
Generative AI at its core is designed as a replacement. to copy.
There are plenty of other solutions that can be implemented with ML. Fun processes are not it.
but this does take away from the creative decisions of doing line art. it’s not you who did the line art anymore.
You find it unreasonable to share my and other’s concerns?
You do you.
I am simply providing the cons of this project. You shouldn’t be taking it to heart.
I’m also not belittling the dev team in any way.
Regarding this, I saw a possibility of this being useful in making concrete lines for filling in, but it has been mentioned that the beta version has a fill tool for gaps in the line art, so why would this feature be needed?