It took me quite a bit to make this video, more than my usual tutorials and it's not even monetized, so yeah it's not that I wanted to just push content, a Pill-format video would be much easier to do... I just think we should think every now and then to the implications of using AI. Looking forward to your feedback.
Matteo, on each new video you post i feel you are not only a graphic designer, a photographer, a programmer, a pillar on AI development, but most of all, a humanist in the very essence of the word. Maybe all the Florence Masters are in your DNA, i feel very humbled by the holistic view you've got. I hope you don't lose your essence to all the greedy, egotistic, un-cooperative society we live in, and if i had the chance to visit Florence, i would definitely love a cup of coffee together. A big thank you from Portugal! PS - i'm an animation director and 3D artist myself, also a "humanist", i would love to contribute somehow, even with my lack of any programming skills, but this video really inspired me. Obrigado.
Hi Mateo, thank you for recall us all on very important topics that we should not forget. Paul here, from France (and expat in Thailand), considered myself still in learning phase, mainly comfyUi and arts in general... Please keep doing what you do. Merci beaucoup 😊!
Matteo , my friend. You are one of the most important persons in this community. Your contributions are undeniable and I'll gladly support you as much as I can via donations or any other means of help. Thank you.
11:40 - If you're not an expert i don't know what counts as an expert. You are definitely qualified. Thank you for your tools, the amazing explanations, and the great Discord community.
Dude, I came for the tutorials for your excellent Comfy Node(s) but after this I'm staying for the nuanced take on AI in our society. I don't really want to do a point by point in the comments thread, but I agree with you more often than I disagree and where our opinions differ it makes me think about my prior assumptions in those areas. Thanks for sharing your thoughts and I look forward to any times in the future you take a break from documentation and technical education to zoom out and talk about the larger more systemic issues around AI and AI use in our world. You rock.
thanks I really appreciate your message. I was very hesitant about this video, but ultimately I don't care much about the likes or number of subscribers so I went for it
Thanks for the great essay! Let everything be art for all I care, much more interesting is what makes good art or bad art. I agree: Making a decent image with AI is very easy now. But making a great one is still very, very hard. It’s just much more difficult now to filter out the bad and appreciate the good ones. But occasionally, you stop and say: oh, right, this is fantastic! It’s just much too early to tell what out of this constant stream of images will last and what will not. --- The environmental implications are a real pain, I must admit I wasn't even thinking about that too much until now.
This is something I've been thinking and talking about. Your video was very insightful, and helped me put into words some things I was struggling to describe. Thanks!
There's a brilliant interview of Dali where he goes on differences, "The only difference between a mad man and Dali is that Dali is not a mad man, the rest is identical" ... Then he stated in his flamboyant way "The only difference between the most perfect photograph in the world and a picture by Velázquez is exactly 7 million dollars". Thanks Matteo for your insight, you rule ❤️
Thank you very much, Matteo, for this video. I believe we should also emphasize the importance of disclosing when AI has been involved in the creation process of any content, as AI differs from other tools by drawing on the creativity of other people. -- Grazie mille, Matteo, per questo video. Credo che dovremmo anche sottolineare l'importanza di segnalare quando l'IA è intervenuta nel processo di creazione di qualsiasi contenuto, poiché l'IA si differenzia dagli altri strumenti attingendo alla creatività di altre persone. -- This message was translated by AI:)
Congratz! ❤ Your content is one of the most enjoyable and educational around!! Thank you so much!! 😊
5 หลายเดือนก่อน +1
Hi, as a quite old (63) German artist, coming from the influence of surrealism and (since the 1990s) fractals, I am allways looking for randomly (aleatoric principle) created artwork, I became a big fan of AI-artwork. For me, I like o play around with the comfy UI possibilities. I just wanna thank You and all the other developers for offer me these new tools. And of course, like every new technology, ther are good and bad implementations of using it. It mostly depends on the ethical intention of the user. That' all I like to say for today, especially my english is not so good. Love and greetings from Lower Saxony...
Thanks for this video. The perspective about ai and how it fits in with art, ethics and environment is an important consideration and having a better understanding makes us better artists!
As a front-end engineer, I completely understand your sentiment and agree with everything you said in this video. You are an incredible person for doing what you do and bringing light to these topics and issues in such a great and non-abrasive way. There are so many times that I wish i could do more to help but I don't know much about Python or the science behind AI that with my toolset I don't know how to help, but honestly if I did know of a way to help or if someone could nudge me in the right way that I could help. i would contribute in a heart beat! Maybe it's time for me to pick up a new skill and learn something new haha, but Like everyone else has already said - we appreciate you and your work tremendously! It really is an inspiration
I really agree with you on most of your statements. Especially the photography metaphor is something I like to use as well. First time I heard it from Adobe's Rufus Deuchler who lives in Florence as well. By the way, Tuscany's artistic heritage is very obvious in your work and your ideas. Thanks for the tools and your knowledge, Matteo. Grazie mille! Sei veramente grande!
"Is AI art?" AI-generated images/pictures/art is definitely art! Even though it uses other people's styles, that doesn’t mean it isn’t valid. Throughout history, many artists have been inspired by others. Movements like Impressionism and Cubism built on existing styles. AI, trained on vast datasets, continues this tradition, creating unique combinations and interpretations. Plus, art is about creativity and expression, and AI-generated pieces can evoke emotions and provoke thought just like human-made art. Think of AI as a new tool or medium, like the camera or computer. If an artist uses AI to create something meaningful, it’s art. It's all about the intent and the creative process. So yes, AI art is real art, thats my opinion.
I have a desktop image by William Turner - Norham Castle, Sunrise (Château de Norham, lever de soleil) . It's a picture he drew after he was photographed with a camera. As shown in the film, he thought it was the end of painting and bequeathed all the paintings to England free of charge. Then came cameras computers computers tablets and 3D graphics. I'm a former 3d artist, gave up sculpting and texturing rigging and rendering as soon as I learnt about the possibilities of automac 1111. I also like to draw on a tablet and I'm into photography, but it was logical to give it all up to learn Python and now I've been practising ComfyUI every day for half a year. I'm trying to master the tools to express my emotions feelings and vision with AI. Of course I think about contrasts lighting and colour in my work and in general I use 3D base and CN on top of that. In conclusion I will add. Sort the works on CivitAi for all time by the number of reactions. Pay attention to the top works, which are no worse than Salvador Dali. They have all the hormones that are artistically required for the joy of the soul and the eye. And my mum also likes the top generations. She doesn't even know what a generative AI image is and she doesn't care how it is made, she likes it in fact. Another more important question is whether a person created a masterpiece by accident or by talent and skill.
I totally agree with you. Always cool to see someone else with an art/design background have a take on this. When I first started with photography, I was often told that it was not art as well and many people still seem to think that it consists of just pointing a camera and pressing a button. One of the reasons that I enjoy ComfyUI so much is because I can use it in my own art workflow (also because I am not impressed with Firefly). I remember first discovering the idea of photo-bashing or matte painting in Photoshop magazine years ago, and how blown away I was by the examples. I hope that as AI matures (and the ideas around it), we will start seeing it the same way.
I was asked if I use AI in my own art recently. I said yes, to get inspiration. But then I went on. I'll post here what my response was since it addresses the same group of questions. All of us here know that eventually, AI will be good enough that you won't even need all these tricks and just ask for what you want, so it is going to get easier, and so more accessible to everyone to get what they're looking for. That response: "Here’s a scenario. You are a creative person. Maybe you’re not the best at whatever you choose to do, or maybe you’re a badass. This scenario doesn’t care which one you are, just that you want to create. Now, let’s say the year is 1984, and you are very busy. Let’s say it’s because you’re traveling a lot to sell some of your paintings you’ve done over the last 10 years. Now, we’ve established that you are a creative person, that you are a painter, and that this is 1984. Now, let’s say you haven’t had time to do creative things lately because you’ve been on the road for the last year. You want to continue your output however, so you hire someone else to learn your style and paint your paintings (you can train AI on your style). You’ve made some sketches (these can be fed into AI with the prompt) and handed them over with instructions (a prompt) on how you want this to turn out, and you pay this person some fee per month (subscriptions) to do these as you come up with ideas. Flash forward another year and this person has output 35 paintings. you pick 20 of them and sign your name on them, then take them to the galleries and the festivals and have a great time selling them as your art. It looks like your art. Other people think it’s your art. Is it your art?" So, of course, you can see from this all the tangles AI is going to bring to the creative sphere, whether it's painting, music, writing, whatever... it will get good enough eventually that you'll be able to ask for something, and it will output it as well as most "good at it" people would be able to do, and even many of those people will likely offload work to AI (i. e. not just muggles using it). When that happens, will we call it art just because an "artist" typed the prompt? Food for thought. BTW, this scenario actually happened in 1984.
Maybe Art will become more of a creating process. If you just want to consume Art you can use AI and have amazing outcome. So the only reason to create art, is because you actually love and enjoy the process. Can't wait to see what the future holds :D
Photoshop itself is not art, but a tool, which can be used to create art. Following the same logic, AI image software could also be seen as a tool being used by an artist. However, it is true that AI art is just not the same. It's much more "composition", than it is drawing/painting. For this reason, I agree with people that say AI generated images should not be protected by copyright. My final point is that Art is in the eye of the beholder. If someone manages to sell a print of "art" they generated with AI, to me there's no question what was sold was art, not a service (i.e. image generation and printing service). But then again, is there truly gonna be a market for art that doesn't have an "original", and that is not protected by copyright? It's going to be a lot more like someone supporting the artist, than someone engaging in Art Collection or Art Speculation.
Matteo, me tomo el atrevimiento de comentar en mi idioma. Decis que no sos un ingeniero, ni tampoco sos un artista. Para mi entonces sos un Mago. Tu magia es increíble, tu aporte es muy valioso!!!. En cada video que subis no dejo de sorprenderme de tu capacidad. Espero que sigas deleitándonos con tu alquimia, te agradezco mucho por compartir tu trabajo!!! Abrazo Grande desde Argentina!!! 🤗
Your are doing a great job with ip adapter. And your videos are top notch ! If it's complicated sometimes, it's just how it is ... comfyui is just hard to master in a short period of time.
I think it's really positive video too see one of the developers of the tech also talking about the downsides and darksides. That being said, the tools and possibilities are amazing and artistry will never be challanged by this tech...it will evovle along side it. As someone working in the VFX industry I can't wait to see the scope of stories we'll be able to tell but also the valuable help we can get to deal with impossible deadlines. Thanks you for all contributions and tutorials!
very well said. A clear and concise presentation of the past history of new technology and art and current issues and concerns from the point of view of an artist and developer. I couldn't agree more.
Thank you Mateo, I agree with you 100%. From day one, I've been as open as possible with the tools I use when I post work online. I think all artists using AI tools have a responsibility to disclose how they're using them in order for the general public to feel secure in their ability to tell when it's being used or not. Otherwise we'll live in a world where nobody trusts anything. Another reason I disclose my process is so people can begin to recognize exactly what the creative process is, and differentiate between simply making a cool image and making an expression meant to communicate a personal experience. Really appreciate your work and the thought you put into it. I'm hoping those of us with good intentions will be able to carry on for the long term in the face of corporate and profit driven interests.
Thank you for this video. This was useful for many of us and especially those are new. Keep making these stuff too. By the way, the gold glittering photo was amazing. Congratulations for 20K. God bless
It's a real Manifesto Matteo. You're a wize man... Thanks for this video and all of you are doing for the community: Your work is fundamental on every aspects !
Matteo, could you please make a deeper dive into what conditioning is? Your tutorials are already great but they're still dodging the rabbit hole of the topic. The thing is, conditioning somehow contains the meaning of the image in CLIP space, while also containing spacial regionality (therefore, different conditionings can be applied to different parts of the same image at once), as well as somehow being able to preserve the meaning of a prompt which exceeds token limit (by splitting it into parts and somehow applying them one after another) and also capable of having various versions of itself on various sampling steps. All of this combined - and you don't really know what you're doing UNLESS you truly understand what conditioning is. Without it, you can't combine advanced techniques together (like, conditioning regionality combined with UltimateUpscale).
I would love to discuss more about conditioning and prompt/ generative regionality as I have always wondered where I was going wrong with my attempts to contain, for example, prompt bleed, and could never really nail down how to control specific aspect generation within the images I have been fortunate enough to generate. If you have the time to explain more, I would love having a chat over a cup of coffee or something!
my two "basics" videos talk a lot about conditioning. but it's a very complex topic and it depends also on how the model is trained, so a generic video is not easy
@@latentvision Thank you for replying! I understand what you mean... Not all models are trained in the same manner, nor extent, so a generic video would be difficult. What about on a suggested/ most popular model? Would it not be accurate to assume that the most popular models are trained using the same, if not very similar, techniques? I will take another look into your ''basics'' videos. Last night I tried your approach to regional control via your fantastic IPAdapter nodes using RGB and separate text prompt boxes, but my PC is not the best (1060 6gb) so I ended up waiting around 5-10 minutes for a single generation - which didn't wok because I think I over indulged in writing the prompts aha
@@latentvision Thanks for the reply. Yeah, I watched your "basics" videos. And re-watched them multiple times now. And yeah, I realize how complex the topic is. Since so far, even after extensive googling, watching yours and other people's tutorials on what happens under the hood, I still have a very vague understanding of what conditioning actually is. It's a mysterious array of tensors which SOMEHOW contains all this data I listed above. The issue is this "somehow" part, which leads to magical thinking and therefore unpredicztable results. Also, I must say that out of all the sources I checked, your tutorials still haven't clarified the subject completely, but they are the ones which approached it MUCH closer than others. So... yeah. If you decide to dedicate some time to finally close this essential chapter, it would be greatly appreciated. Anyway, thanks for the great work you've already done. You're awesome!
Excellent video. I love your explanation here. I’ve long felt it’s weird there’s even a debate if ai is art. It’s a tool. Who calls a tool “art”? You CAN make art with the tool. But the tool itself is not art.
I believe the debate is more about "prompt engineers" that call themselves artists. They take jobs from actual artists that know what they are doing but are not yet prepared for the AI revolution. My suggestion to artists is to learn the tool and use it if you think it helps you. A person with an artistic background will always be able to get better results from the technology, like a photographer will always be able to get a better photo than the average person with an IPhone.
@@latentvision yep, I heard you mention that in the video.... and yeah, I agree. I did photography for 35+ years. I can tell the lighting issues in most ai photos easily are way off. I also agree that the main argument that matters in art is the prompt engineers stealing jobs you mentioned... but .... that's not where most of the masses are focused. I work in IT and also have a lot of non tech and non artist friends .... 99% of the people around me are still at the point of just debating whether the tool itself is "art" or not. they dont' even know what a "prompt engineer" is. and I suspect that's the common folk too. in fact, in my tech job, senior management - also doesn't know what a prompt engineer is. not disagreeing your point. I fully agree where the metal hits the ground, your point is valid as that's what ultimately counts for art and ethics around the topic. I'm only pointing out that the folks voting and making decisions on the space... have a different optic on the situation (one that's even worse in my opinion)
Thanks bro for such deep video! I agree with your position and believe that we should focus on mindfulness, ethical usage of AI and promote it. It can be very helpful in the right hands. Again, Matteo, Thank you ❤🔥
This was a thought-provoking and well done video. I appreciate you taking the time to step back and think about the bigger picture; and of course I'm deeply thankful for all your work to create and teach SD tools. A few additional thoughts (please forgive the length, but I wanted to be as rigorous and nuanced as possible): I agree that some people will always complain about models that are anything but 100% free and that these complaints are often misplaced. However, I think the bigger issue with Mobius is that it was not clear what that model was actually offering; and, when faced with questions, the creators weren't able to give good answers and instead became outright hostile. Then there was subsequent evidence that it might have actually have been more of a scheme to boost a crypto project than any sort of actual breakthrough when it comes to new and useful models. As far as energy use, it's important to first keep in mind that these are just projections. But even if we assume they turn out to be correct, we need to apply some perspective. 4% of global electricity use equates to (at most) about 1% of global GHG emissions. Not nothing, but still small compared to things like driving, flying, and even air conditioning. If people are inclined to focus on the small contribution of AI before tackling these other bigger things, their issue is more with AI than with climate change. Another important consideration is the fact that data centers are often colocated with renewable generation, so associated emissions are probably lower for this particular use. All this said, the solution remains the same: more renewable electricity generation so that we decarbonize electricity, no matter what we're using it for. I could say even more, but just want to finally note that the trash "island" is the Pacific is a common misunderstanding of something very real. It's true there is an area of concentrated plastic waste in the Pacific, but contrary to popular belief, it's not an island, most of it is microplastics, most of what is not microplastics is fishing nets and fishing gear, and about 20% of it (as of 2018) was leftover from the Japanese tsunami. It's truly mindblowing to read about: education.nationalgeographic.org/resource/great-pacific-garbage-patch-isnt-what-you-think/
I agree on everything you said. Regarding the "island" I was reading the article as it was written. It's actually a "patch" and it's estimated to contain 100,000 metric tons of garbage (the article you posted is unfortunately outdated). "island trash" is more sensationalist but it doesn't mean that it's not there. I would also like to point out that the one in Hawaii is just one the 5 patches of garbage in the oceans.
Thanks for all you are doing Mattéo. 😇 If I may share my opinion : I think Art is in the eye of the beholder. So If someone sees Art in your creations, then you're an artist, whatever tools you are using. Contemporary art teaches us to free ourselves from technique.
defining art is not easy and I don't pretend to know the answer. generally speaking art requires "intention", it has to have a cultural or historical relevance, it needs to transmit a message... and even the lack of message is a message. Exercising visual arts doesn't mean that you are making Art. Something can be beautiful for you and you can tell yourself that it's art and that would be 100% true for you, but if you show it to some friends and they say it's not art for them. So what is it? Their vision is as good as yours. If we are talking about Art (capital) we need some kind of importance for the betterment of humanity
This is a great video Matteo. I have a different point of view when you refer to art, but it's refreshing seeing opinionated discussion on this topic that isn't charged or evangelizing the viewer. I feel like much of the way creators can monetize their work through platforms such as youtube, patreon etc incentivize tribalism and hinder collaboration if the creators want to get the most out of their effort. It would be great having an option such as foldingathome, where individuals can put their hardware towards community-driven collaborative training efforts. I think that might happen in the distant-ish future and will be a significant advantage for local models, allowing them to be trained without relying on companies that might drop open source at any moment. Hopefully no future legislation gets in the way.
Amazing video. I put off watching it because the "Is AI art?" topic usually devolves into a misunderstanding of how AI works from traditional artists and disrespect or disregard of the vast history and importance traditional artists have contributed to culture from the AI users. I appreciate seeing an argument being made that has respect for both sides with an understanding of each perspective. This is an insightful video, thank you for making it.
Well, it is not easy to find someone you totally agree with everything they say, and you're not an exception. Don't sell yourself short. Your dedication and efforts make you one of the most important individuals in the world of AI. That was my opinion, and today, with your insight and talk about ethics and controversies of the use of AI and its future, I am more convinced.
Я живу в самой зеленой стране на свете. Где 50% вырабатываемой энергии уже зеленая(атомная и гидро-энергетика), а остальные потребители находятся так далеко от центральной части моей страны, что огромная доля энергии попросту расходуется в ничто. Так что да. Я могу генерировать сколько угодно вайфу. Не беспокойся, Матео. Ты делаешь крутые вещи!
I have said it many times before and I'm going to keep saying it until everyone is saying it. ALL models should be open source. It should be the law that they have to be. They are created from us they would not exisist without us they should belong to all of us. *We need to fight for this.* Otherwise what he says in this video will be true. All AI will end up belonging to large corperations which is exactly what they want. Corpos don't want us to have this power without paying them for it, even though its build off the backs of every person on the planet.
Thanks Matteo for your videos and thoughts. Those of us who have been around for a few decades recognise the repeated pattern of condemning new tools as they are perceived as threats to artistic integrity. As you point out Photography is an obvious one. I lived through the introduction of personal computers and the condemnation from artists the moment people started using them to produce images. Now of course a vast number of artists use computers exclusively to produce their art. AI or the algorithms behind AI are of course more nuanced and complex but I agree are still a 'human made tool' that can be used to create artistic expressions. It nearly always comes back to who is behind the tool and what is their intent. Already I see creative people making wonderful images that stand out from the templated button pushers. The fusion of traditional, digital and AI tools is definitely one way forward. Just looking at art history I have little doubt a new expression of art will be born where the use of AI is central and eventually accepted by the majority. It might just take a few decades...🙂
I think you are wrong on this one. What is art? Art is what conveys feelings. What is Art is decided by the audience, not by the artist. If you see a scene of a homeless guy with his cart under the rain, you feel something, yet there is no director behind it. So, can generative AI produce art? Of course, it's all in the audience reaction, in how an image clicks with the watcher. Now, the real question is, is the guy who simply types prompt an artist? No, the artist is the AI, the guy typing prompt can call himself an art director, which is the one ask the artist what to do, and makes a selections on what the artist produce.
Thanks for comments. I took a similar path as you. I was a fine arts major but worked in IT as a career for 30 years because it came natural to me. I still draw and paint and visual artists that make their living this way are right to be concerned. But I also think people who know and enjoy art will always appreciate and want the hand of the artist. Also, AI cannot sculpt yet 😂. Michelangelo felt sculpting was superior to drawing and painting. Just another point of view. Thanks for your great work. I hope you keep going helping us tinkerers.
I share a quite similar background with you, I have a Bachelor degree in photography and I'm a self taught designer, as well as coder. I find it fascinating how AI allows me to connect art and tech in a really novel way. I also share the same feeling that AI is a tool, that can be misused, but can also be used to create wonderful pieces. I'd even go further and say that creating good tools (like the ones you create) is also a form of art, just like we value a good paintbrush made of good materials and we pay more for it.
Thank you for posting a wonderful topic. I really hope that the tools we have today aren't placed behind lock and key. I'll be honest I love art but i am not talented. These tools have granted me the ability to express myself.
I did not expect this man, I completely agree. that was refreshing & I'm glad you didn't give the snake oil tech-bro answer i came to expect from AI channels
I am a trainer and the moment training can't be done locally, for whatever reason, is the moment I uninstall all this. Having said this, and as said trainer, I am sick and tired of nothing being done for the advancement of training. I see some advancements for inference, but we are using the same tools for training I used with 1.4 (under the hood) yet not even the models are the same. Basically, I am trying to write my War and Peace using stone tablets and a few chisels. Not sure why there have been no real advancements for us in training.
@@shisofox All of them, but if you look they are all doing about the same things but no advancements. In inference we have advancements, and stuff we never had in 1.4, but why am I stuck with the same tools, and pipelines, since then for training? I'll tell you why because the companies are buying up the talent that knows how stuff works and could get it done. Most of the people I hung out with, or around, are now working for SAI, for instance, or other companies, so their githubs are either stagnant or shut down.
@@generalawareness101 I'm not nearly as connected as you. The only thing I remember learning about is vision models for generating better captions. It sounds very expensive. Can you ask your former contacts what tools they have that might improve training, that don't cost millions to run?
@@shisofox There is a difference as they are in a business while I am talking about the advancements we have seen in just inference. 1 step gens, 8 step, etc... those are peanuts. Look at IPA for inference. That is a huge advancement for inference. Look at control net, etc... these are all advancements for inference, but we are still using the same training tools we have been. Over on the LLM side people have told me the tools they have they wish they had on SD as they would be phenomenal, but I can only take their word for it as I don't do LLMs. My whole point is the training side is just flat out stagnant, while the inference side has some movement.
Thanks so much your video has given me enough confidence now to publish my ai genersted stuff bevauee i will specially convey my viewers that i am using ai art to convey a msg to them .
I really love that you are so considerate and contribute so positively to the community. Do you have a recommendation of how to get involved with projects to start helping the community? Eg. Start by following this git project or hugging face group and go from there?
if you don't have specific technical skill, I guess the best is helping others resolving basic issues. Check the bug trackers, 90% of the times they are very simple support requests.
Hello Matteo، Your good tutorials changed my perspective on this art, thank you Please do detailed tutorial on controlnets and frame interpolation Thank you in advance
Certainly, incorporating AI into the art-making process is akin to using traditional tools like pencils, brushes, and digital software. Just as artists throughout history have utilized various materials to express their creativity, AI serves as a modern tool for artists to explore and innovate. This fusion of technology and artistry gives rise to a new movement, AI-ISM, where the outcome is not just a mere product but a manifestation of human ingenuity amplified by artificial intelligence.
I agree with you! It was not boring and your point makes total sense. I don't know what makes the difference in the mind that makes you want to collaborate with open source and community. I empathize with your disappointment and I don't know what's the solution, I cheers for some of those contributors become a big company and payback all the unbalance between those who uses and monetize and those who gives something back. Good video
I agree with a lot of what you have said. To me ai is meant to be used as a tool and those of us who learn it's intricacies will be propelled forward. Of course there will be those who will abuse it in undesirable ways, but that can also be said about most anything we already have. In time tech will catch up so we can much easier filter out and find all the ai generated stuff so people don't get fooled when it's used in an exploitative manner (politics etc). Right now it's still the wild wild west in terms of control/regulation/law so of course many ppl will get fooled but i think that in itself is something we have to go through to develop a safer future with a i. It's too much a useful tool to put back in the box, and so much more can be done with it in many industries.
Well its true, we have increadible problems regarding our lack of sustainability. This was a jurney, i started 10 years ago as an awakening. The real Problem is, not climatechange for example, all these problems ar connected to another and not separated Crisises. They are conected through only one single metacrysis. Called Overshoot. So at the moment we use up even our renewable ressources about 70% faster as they can regenerate. like a bankaccount, you could live from for ever if you just used the interest. But that's not what we are doing. We are taking the interest of about 2.7% of the biological capacity of the planet, plus another 1.5%, by eating up the deposit capital, so to speak. The consequences: overfishing, loss of forests, topsoil, biodiversity, etc. And we are not yet talking about non-renewable resources. Especially when it comes to energy and energy resources, we are living in an imaginary fairyland. The only reason oil has not peaked is because fracking in the USA prevented production from peaking for 15 years. But this will not happen again. Alternatives are not in sight fast enough. It will be a bumpy ride into adulthood for humanity. With the realization that limits are real and trees don't grow to the sky. And AI won't change that. For those of you who would like learn about these topics from one of our greatest minds, Prof. William Rees, who was one creator of the ökological footprint th-cam.com/video/LQTuDttP2Yg/w-d-xo.html And for this channel a talk about Ai and Sustainability from Nathe Hagens and Daniel Schmachtenberger.. very deep, but excessive long, but worth it.. th-cam.com/video/_P8PLHvZygo/w-d-xo.html (one impressive center statement from the whole interview above. if 3 hour are to long... just start here. but its worth the effort as well as the ökological superhero william rees in the first link th-cam.com/video/_P8PLHvZygo/w-d-xo.htmlsi=Sn9LnmTbwHDAk0VB&t=9946).
Hi Matteo, thanks for your work!!! May i introduce Nicéphore Niépce, He is the inventor of photography in 1827, then called “heliographic process” the invention of a new industrial print "multiple": this makes sense with the evolution of computer neural network technology. Maybe not yet an art
As a traditional artist, I am not so concerned with the philosophical questions of whether AI art is art (although I don't think that discussion is invalid), but rather with the virulent resistance of some institutionalized artists to this powerful tool and its possibilities. Beyond the ideological biases in this resistance, there is also a lot of snobbery, a belief in being very special individuals. It would be unfortunate if, trying to prevent their style or work from being "stolen," they instigate the imposition of laws that aim to limit the use of generative AIs, but end up favoring large tech companies more than protecting artists' turf.
Agreed, although I'd still feel much better about using these tools, if living artists who didn't give their concent would be excluded from the training data and those who did, would get some compansation. I avoid deliberately prompting for living artists for everything I share publicly, but that approach of course only gets you so far.
Attention AI companies: Hire this man to guide your model development, sit on the ethics board, and go on TV as the face of your tech. And Pay him a lot to just code!
I think the key element of art makes you feel something is that it must be an intentional expression. In this way an autogenerated piece of imagery doesn't necessarily count as art even if it stirs and inspired the viewer. If however someone chose the location to present the autogenerated piece and timing to show it, then it becomes part of a performance art piece. The key is people intend to express something, which is the artist feeling something and using the art medium of choice to do so.
You forget that AI only derives its creations from already intentional human art. Even a drawing style has an original intention (ex, cubism) and is called an artistic style, which is encompassed within the representations of human art. And this would count even ignoring the fact that the person who asks the AI can provide the intention. And that ignores the fact that there are infinite ways to create with AI beyond just writing an order and that's it.
@@cesar4729 i didn't forget at all. i simply stuck to the point i was trying to make. My perspective is that what a person makes with their chosen medium, regardless of the tools used, is an intention to express something they feel which is what makes it art.
@@magejoshplays That is too rigid as a definition. It means that if I make a sculpture for no reason, no one has the right to find artistic value in it. Intention cannot be measured, nor can an author from other centuries say if he did something with any intention. The world does not work that way. It is the observers who decide what to call art or not, and it is the majority collective consensus that allows a relative definition of art in each culture and era. There is *no* concrete definition, since it is a subjective and fluid concept. The closest definition is that art is what people consider art to be in that period.
Say you are an artist and you train a model over your art. Then you use that model to speed up or augment your work, say to make backgrounds. Then you improve the generations by retouching them, adding elements, fixing colors. AI becomes part of a process and in that case in my opinion completely transparent. That in my book is a good use of the technology. Art can be many things but I mostly agree that needs to be intentional. Even though it's not always the case. It can be defined art by "history" or for its cultural relevance. Think for example of the "Moka pot", it was not intended as art but it's now showcased at the Moma
@@cesar4729 I'd suggest that even if you label it as no reason, you still in fact had reason and it therefore becomes art. Then I'd suggest that if someone else comes along and decides they interpret meaning from that expression, they have also defined it as art even if not consciously. Because, as you suggested, ultimately all meaning and definition is entirely relative to the perceiver and all arguments otherwise are simply a continuation of the argument of sovereign soul or subject to a master.
Very interesting and thoughtful. As a professional painter and illustrator AI has been a fascinating journey for nearly a year. I still don't know what my feelings are. Maybe one indicator is that I would never sign a Comfy generation. I don't feel that I made them. Perhaps all images are somewhere in the code and we are explorers seeking them, rather than artists creating them. My painter friends see it as a threat, but no one will want an AI image framed on the wall, sure commercial artists are going to feel a great deal of pain, but people will still want hand created items in their homes. I do wonder what worth the tsunami of images has, no-one gives most of them more than the briefest glance. Is their value in the brief moment of pleasure the creator feels? Maybe, there seems to be more interest in workflows than the end result. When I post videos I only get comments about how they were made and almost never about the images themselves.
Yeah, so far it feels a lot like building sand castles, being proud of them for a little while and then have them washed away by the waves the next day. I'm usually more proud of my workflows than of the images they produce. But as soon as I find a real professional usecase for all of this, maybe that feeling will change.
In my opinion, Art is not the media, the Technik or the material used to create something. Of course mastery used crafrmansship for what ever shows the love someone puts in his efforts and this creates things of wealth and beauty on its own. But areal piece of art for me is something else. It needs to describe a thing a feeling a thought in a way that was never told before. It should enrich our point of views about this world to get a bigger view of reality or to say enlarges also the reality of mind.
We really need these additional tools. Language is fine for AI dialogue generation, but it is too scarce for images. In reality, it is difficult for us to use language to carefully annotate a picture, and then apply this annotation method to AI painting. It is difficult or impossible to achieve the picture you imagine simply through prompt words. You can only use Lora or ControlNet to tell AI the desired picture effect through picture to picture. That means you must have pictures first to generate the imagination in your mind. It is a very painful logic. Will there be something beyond prompt words in the future?Connect my brain to controlnet!
Thank you Matteo for your hard work! In the first part of the video, you mentioned competition and what impetus it gave to the development of art. Unfortunately, generative technologies have been taken out of the competitive field. There is still no clear idea who is the author of the generated image. Is it possible to use it for commercial purposes? There is also no legal assessment given for collecting material for training models. All this rather slows down the development of art.
Imagine all AI companies merging their LLMs and Generative art models, we are affraid that AI will outgrow us but as long as these companies keep competing with eachother training their own models with 70 billion and 400 billion and what not parameters, think of what a merge between Llama3 and OpenAI would cause, that merger would blow EVERY other existing model out of the sky and AGI would probably arrive a lot faster than expected, but as long as they keep competing AI will save no one and the energy consumption will grow as is
I only disagree ( and not completely) on prompt engineering. Yes it does not make you an artist, as much as knowing how to get the best out of a DSLR camera does not make you an artist, but it sure helps out a lot. And it makes you more proficient and capable of getting art out of your tool. It's a skill that contributes to your art. I agree on everything else. If, as a community, we could rebuild insightface training datasets and re-train on it to release open source weights it would be an achievement. Thanks for all your work
I'm not saying that "prompt engineering" is not important. I'm saying that it's not engineering and doesn't make you an artist. Also it's a term invented by the AI companies so they can justify the loss of jobs over AI... because they give work to "prompt engineers"
@@latentvision I'm not entirely sure about it's definition. I believe it's real application is in alignment and model weight "exploration". It is used to try and understand the model propagation, limits, and strong points. And sure enough is not how the term is used normally, we might be more in the field of prompt design. It's a big AI company term because only them can really put engineering behind datasets. Yet lllyasviel recent Omost still uses a form of prompt engineering and canvas control to achieve quite interesting results. I feel like I'm debating on terminology just for the sake of it, I'm sorry 😅
Sooner rather than later, AI will be so deeply integrated into countless areas that all the current complaints against AI-generated art will age poorly.
Artists will call a red sqare painted over a green background Art, and then freak out at AI created images. Jokes asides, I find the "Is this art" debate very boring. I find the AI negationists dull and annoying. If AI is a tool, whatever you produce with it could be art. Same goes for a brush, photoshop, a guitar or a mixing board. Quoting Robert Fripp: "I don't like that word, Art. If one sees art as that breeze that bloweth as it listeth that intangible quality, that certain something, the magic of a performance... we can't control that. But we can raise a sail to catch the breeze." And here we are, debating about the cloth of the sail.
First of all many thanks for all the work done and for the tutorials! Then i guess that gen ai is a bit like ebikes: if I use one I can make 50 km in a day, if you give it to a pro they can make 500 km in a day. So, I would say, if you want to make art with ai you'd better know some colour theory and composition, or if you want to do brand communication you'd better know some commercial art direction. The same applies to language models. Last but not least: are plastic testicles good with sex dolls? 🤔
A scribble by a 3-year-old is still art. It's a unique product that was created over time by applying skill and imagination. It might not be art that will impress many people in terms of technicality or innovation, but that's not obligatory.
It took me quite a bit to make this video, more than my usual tutorials and it's not even monetized, so yeah it's not that I wanted to just push content, a Pill-format video would be much easier to do... I just think we should think every now and then to the implications of using AI. Looking forward to your feedback.
👋
Congrats on your subscriber milestone 👍
Great video. Thank you for the sound words.
Matteo, on each new video you post i feel you are not only a graphic designer, a photographer, a programmer, a pillar on AI development, but most of all, a humanist in the very essence of the word. Maybe all the Florence Masters are in your DNA, i feel very humbled by the holistic view you've got. I hope you don't lose your essence to all the greedy, egotistic, un-cooperative society we live in, and if i had the chance to visit Florence, i would definitely love a cup of coffee together. A big thank you from Portugal! PS - i'm an animation director and 3D artist myself, also a "humanist", i would love to contribute somehow, even with my lack of any programming skills, but this video really inspired me. Obrigado.
Hi Mateo, thank you for recall us all on very important topics that we should not forget.
Paul here, from France (and expat in Thailand), considered myself still in learning phase, mainly comfyUi and arts in general...
Please keep doing what you do. Merci beaucoup 😊!
Matteo , my friend. You are one of the most important persons in this community. Your contributions are undeniable and I'll gladly support you as much as I can via donations or any other means of help.
Thank you.
This,
aaw thanks! I'm not worthy
@@latentvisionDon't say such things my boy. You are ❤
11:40 - If you're not an expert i don't know what counts as an expert. You are definitely qualified. Thank you for your tools, the amazing explanations, and the great Discord community.
I was on your Discord a while back and even then you were helping many users and that is a good thing. Glad to see it expand to help even more people.
Excellent Video Matteo!!
And congratz on your 20k!
Just when I thought Matt3o couldn't possibly do anymore, he drops this amazingly-well thought out piece. I really appreciate these insights.
Dude, I came for the tutorials for your excellent Comfy Node(s) but after this I'm staying for the nuanced take on AI in our society. I don't really want to do a point by point in the comments thread, but I agree with you more often than I disagree and where our opinions differ it makes me think about my prior assumptions in those areas. Thanks for sharing your thoughts and I look forward to any times in the future you take a break from documentation and technical education to zoom out and talk about the larger more systemic issues around AI and AI use in our world. You rock.
thanks I really appreciate your message. I was very hesitant about this video, but ultimately I don't care much about the likes or number of subscribers so I went for it
Thanks for the great essay! Let everything be art for all I care, much more interesting is what makes good art or bad art. I agree: Making a decent image with AI is very easy now. But making a great one is still very, very hard. It’s just much more difficult now to filter out the bad and appreciate the good ones. But occasionally, you stop and say: oh, right, this is fantastic! It’s just much too early to tell what out of this constant stream of images will last and what will not. --- The environmental implications are a real pain, I must admit I wasn't even thinking about that too much until now.
Love your explanation on this topic. Congrats on the 20k subs! 🎉🎉🎉🎉
Thank you for your video Matteo, I love your stuff.
I know that this is not your usual content, and that's why I find it so interesting to hear your opinion. Thanks for sharing!
This is something I've been thinking and talking about. Your video was very insightful, and helped me put into words some things I was struggling to describe. Thanks!
There's a brilliant interview of Dali where he goes on differences, "The only difference between a mad man and Dali is that Dali is not a mad man, the rest is identical" ... Then he stated in his flamboyant way "The only difference between the most perfect photograph in the world and a picture by Velázquez is exactly 7 million dollars".
Thanks Matteo for your insight, you rule ❤️
Thank you very much, Matteo, for this video. I believe we should also emphasize the importance of disclosing when AI has been involved in the creation process of any content, as AI differs from other tools by drawing on the creativity of other people.
--
Grazie mille, Matteo, per questo video. Credo che dovremmo anche sottolineare l'importanza di segnalare quando l'IA è intervenuta nel processo di creazione di qualsiasi contenuto, poiché l'IA si differenzia dagli altri strumenti attingendo alla creatività di altre persone.
--
This message was translated by AI:)
Congratz! ❤ Your content is one of the most enjoyable and educational around!! Thank you so much!! 😊
Hi, as a quite old (63) German artist, coming from the influence of surrealism and (since the 1990s) fractals, I am allways looking for randomly (aleatoric principle) created artwork, I became a big fan of AI-artwork.
For me, I like o play around with the comfy UI possibilities.
I just wanna thank You and all the other developers for offer me these new tools.
And of course, like every new technology, ther are good and bad implementations of using it.
It mostly depends on the ethical intention of the user.
That' all I like to say for today, especially my english is not so good.
Love and greetings from Lower Saxony...
thanks for taking the time write. appreciate it
Thanks for this video. The perspective about ai and how it fits in with art, ethics and environment is an important consideration and having a better understanding makes us better artists!
Love all your work! Thanks for the sharing!
As a front-end engineer, I completely understand your sentiment and agree with everything you said in this video. You are an incredible person for doing what you do and bringing light to these topics and issues in such a great and non-abrasive way. There are so many times that I wish i could do more to help but I don't know much about Python or the science behind AI that with my toolset I don't know how to help, but honestly if I did know of a way to help or if someone could nudge me in the right way that I could help. i would contribute in a heart beat! Maybe it's time for me to pick up a new skill and learn something new haha, but Like everyone else has already said - we appreciate you and your work tremendously! It really is an inspiration
I really agree with you on most of your statements. Especially the photography metaphor is something I like to use as well. First time I heard it from Adobe's Rufus Deuchler who lives in Florence as well.
By the way, Tuscany's artistic heritage is very obvious in your work and your ideas.
Thanks for the tools and your knowledge, Matteo.
Grazie mille! Sei veramente grande!
YOU ARE AMAZING. Please keep going!
"Is AI art?"
AI-generated images/pictures/art is definitely art! Even though it uses other people's styles, that doesn’t mean it isn’t valid. Throughout history, many artists have been inspired by others. Movements like Impressionism and Cubism built on existing styles. AI, trained on vast datasets, continues this tradition, creating unique combinations and interpretations. Plus, art is about creativity and expression, and AI-generated pieces can evoke emotions and provoke thought just like human-made art. Think of AI as a new tool or medium, like the camera or computer. If an artist uses AI to create something meaningful, it’s art. It's all about the intent and the creative process. So yes, AI art is real art, thats my opinion.
I absolutely agree.
I have a desktop image by William Turner - Norham Castle, Sunrise (Château de Norham, lever de soleil) . It's a picture he drew after he was photographed with a camera. As shown in the film, he thought it was the end of painting and bequeathed all the paintings to England free of charge. Then came cameras computers computers tablets and 3D graphics. I'm a former 3d artist, gave up sculpting and texturing rigging and rendering as soon as I learnt about the possibilities of automac 1111. I also like to draw on a tablet and I'm into photography, but it was logical to give it all up to learn Python and now I've been practising ComfyUI every day for half a year. I'm trying to master the tools to express my emotions feelings and vision with AI. Of course I think about contrasts lighting and colour in my work and in general I use 3D base and CN on top of that. In conclusion I will add. Sort the works on CivitAi for all time by the number of reactions. Pay attention to the top works, which are no worse than Salvador Dali. They have all the hormones that are artistically required for the joy of the soul and the eye. And my mum also likes the top generations. She doesn't even know what a generative AI image is and she doesn't care how it is made, she likes it in fact. Another more important question is whether a person created a masterpiece by accident or by talent and skill.
I totally agree with you. Always cool to see someone else with an art/design background have a take on this. When I first started with photography, I was often told that it was not art as well and many people still seem to think that it consists of just pointing a camera and pressing a button. One of the reasons that I enjoy ComfyUI so much is because I can use it in my own art workflow (also because I am not impressed with Firefly). I remember first discovering the idea of photo-bashing or matte painting in Photoshop magazine years ago, and how blown away I was by the examples. I hope that as AI matures (and the ideas around it), we will start seeing it the same way.
I was asked if I use AI in my own art recently. I said yes, to get inspiration. But then I went on. I'll post here what my response was since it addresses the same group of questions. All of us here know that eventually, AI will be good enough that you won't even need all these tricks and just ask for what you want, so it is going to get easier, and so more accessible to everyone to get what they're looking for.
That response:
"Here’s a scenario. You are a creative person. Maybe you’re not the best at whatever you choose to do, or maybe you’re a badass. This scenario doesn’t care which one you are, just that you want to create. Now, let’s say the year is 1984, and you are very busy. Let’s say it’s because you’re traveling a lot to sell some of your paintings you’ve done over the last 10 years.
Now, we’ve established that you are a creative person, that you are a painter, and that this is 1984. Now, let’s say you haven’t had time to do creative things lately because you’ve been on the road for the last year. You want to continue your output however, so you hire someone else to learn your style and paint your paintings (you can train AI on your style). You’ve made some sketches (these can be fed into AI with the prompt) and handed them over with instructions (a prompt) on how you want this to turn out, and you pay this person some fee per month (subscriptions) to do these as you come up with ideas.
Flash forward another year and this person has output 35 paintings. you pick 20 of them and sign your name on them, then take them to the galleries and the festivals and have a great time selling them as your art. It looks like your art. Other people think it’s your art.
Is it your art?"
So, of course, you can see from this all the tangles AI is going to bring to the creative sphere, whether it's painting, music, writing, whatever... it will get good enough eventually that you'll be able to ask for something, and it will output it as well as most "good at it" people would be able to do, and even many of those people will likely offload work to AI (i. e. not just muggles using it). When that happens, will we call it art just because an "artist" typed the prompt? Food for thought.
BTW, this scenario actually happened in 1984.
Maybe Art will become more of a creating process. If you just want to consume Art you can use AI and have amazing outcome. So the only reason to create art, is because you actually love and enjoy the process. Can't wait to see what the future holds :D
I 1000 percent agree
Yes indeed, you are right. Many thanks for producing such a great video!
Photoshop itself is not art, but a tool, which can be used to create art.
Following the same logic, AI image software could also be seen as a tool being used by an artist.
However, it is true that AI art is just not the same. It's much more "composition", than it is drawing/painting. For this reason, I agree with people that say AI generated images should not be protected by copyright.
My final point is that Art is in the eye of the beholder. If someone manages to sell a print of "art" they generated with AI, to me there's no question what was sold was art, not a service (i.e. image generation and printing service). But then again, is there truly gonna be a market for art that doesn't have an "original", and that is not protected by copyright? It's going to be a lot more like someone supporting the artist, than someone engaging in Art Collection or Art Speculation.
Matteo, me tomo el atrevimiento de comentar en mi idioma. Decis que no sos un ingeniero, ni tampoco sos un artista. Para mi entonces sos un Mago. Tu magia es increíble, tu aporte es muy valioso!!!. En cada video que subis no dejo de sorprenderme de tu capacidad. Espero que sigas deleitándonos con tu alquimia, te agradezco mucho por compartir tu trabajo!!! Abrazo Grande desde Argentina!!! 🤗
thanks for the kind words
I am addicted to Comfiui.
Your are doing a great job with ip adapter. And your videos are top notch ! If it's complicated sometimes, it's just how it is ... comfyui is just hard to master in a short period of time.
Grazie Matteo ❤ per la tua 'Consistent Vision' e per l'impegno che hai messo in tutto il tuo lavoro, ne apprezzo il valore e la lungimiranza ❤❤❤
I think it's really positive video too see one of the developers of the tech also talking about the downsides and darksides. That being said, the tools and possibilities are amazing and artistry will never be challanged by this tech...it will evovle along side it. As someone working in the VFX industry I can't wait to see the scope of stories we'll be able to tell but also the valuable help we can get to deal with impossible deadlines.
Thanks you for all contributions and tutorials!
Thank you, Matteo, for your continuous efforts.😊
Matteo , my friend. You are one of the most important persons in this community. I really appreciate these insights
very well said. A clear and concise presentation of the past history of new technology and art and current issues and concerns from the point of view of an artist and developer. I couldn't agree more.
Thank you Mateo, I agree with you 100%.
From day one, I've been as open as possible with the tools I use when I post work online. I think all artists using AI tools have a responsibility to disclose how they're using them in order for the general public to feel secure in their ability to tell when it's being used or not. Otherwise we'll live in a world where nobody trusts anything.
Another reason I disclose my process is so people can begin to recognize exactly what the creative process is, and differentiate between simply making a cool image and making an expression meant to communicate a personal experience.
Really appreciate your work and the thought you put into it. I'm hoping those of us with good intentions will be able to carry on for the long term in the face of corporate and profit driven interests.
Матео, спасибо тебе большое за то что ты делаешь!
И я согласен с тобой полностью в твоем виденье ИИ и этики работы с ним.
Great point, artists really do all need to think independently and diversify.
Thank you for this video. This was useful for many of us and especially those are new. Keep making these stuff too. By the way, the gold glittering photo was amazing. Congratulations for 20K. God bless
It's a real Manifesto Matteo. You're a wize man... Thanks for this video and all of you are doing for the community: Your work is fundamental on every aspects !
Matteo, could you please make a deeper dive into what conditioning is?
Your tutorials are already great but they're still dodging the rabbit hole of the topic.
The thing is, conditioning somehow contains the meaning of the image in CLIP space, while also containing spacial regionality (therefore, different conditionings can be applied to different parts of the same image at once), as well as somehow being able to preserve the meaning of a prompt which exceeds token limit (by splitting it into parts and somehow applying them one after another) and also capable of having various versions of itself on various sampling steps.
All of this combined - and you don't really know what you're doing UNLESS you truly understand what conditioning is. Without it, you can't combine advanced techniques together (like, conditioning regionality combined with UltimateUpscale).
I would love to discuss more about conditioning and prompt/ generative regionality as I have always wondered where I was going wrong with my attempts to contain, for example, prompt bleed, and could never really nail down how to control specific aspect generation within the images I have been fortunate enough to generate. If you have the time to explain more, I would love having a chat over a cup of coffee or something!
my two "basics" videos talk a lot about conditioning. but it's a very complex topic and it depends also on how the model is trained, so a generic video is not easy
@@latentvision Thank you for replying! I understand what you mean... Not all models are trained in the same manner, nor extent, so a generic video would be difficult. What about on a suggested/ most popular model? Would it not be accurate to assume that the most popular models are trained using the same, if not very similar, techniques? I will take another look into your ''basics'' videos. Last night I tried your approach to regional control via your fantastic IPAdapter nodes using RGB and separate text prompt boxes, but my PC is not the best (1060 6gb) so I ended up waiting around 5-10 minutes for a single generation - which didn't wok because I think I over indulged in writing the prompts aha
@@latentvision Thanks for the reply.
Yeah, I watched your "basics" videos. And re-watched them multiple times now. And yeah, I realize how complex the topic is. Since so far, even after extensive googling, watching yours and other people's tutorials on what happens under the hood, I still have a very vague understanding of what conditioning actually is. It's a mysterious array of tensors which SOMEHOW contains all this data I listed above. The issue is this "somehow" part, which leads to magical thinking and therefore unpredicztable results.
Also, I must say that out of all the sources I checked, your tutorials still haven't clarified the subject completely, but they are the ones which approached it MUCH closer than others.
So... yeah. If you decide to dedicate some time to finally close this essential chapter, it would be greatly appreciated. Anyway, thanks for the great work you've already done. You're awesome!
Excellent video. I love your explanation here. I’ve long felt it’s weird there’s even a debate if ai is art. It’s a tool. Who calls a tool “art”? You CAN make art with the tool. But the tool itself is not art.
I believe the debate is more about "prompt engineers" that call themselves artists. They take jobs from actual artists that know what they are doing but are not yet prepared for the AI revolution.
My suggestion to artists is to learn the tool and use it if you think it helps you. A person with an artistic background will always be able to get better results from the technology, like a photographer will always be able to get a better photo than the average person with an IPhone.
@@latentvision yep, I heard you mention that in the video.... and yeah, I agree. I did photography for 35+ years. I can tell the lighting issues in most ai photos easily are way off.
I also agree that the main argument that matters in art is the prompt engineers stealing jobs you mentioned... but .... that's not where most of the masses are focused. I work in IT and also have a lot of non tech and non artist friends .... 99% of the people around me are still at the point of just debating whether the tool itself is "art" or not. they dont' even know what a "prompt engineer" is. and I suspect that's the common folk too. in fact, in my tech job, senior management - also doesn't know what a prompt engineer is.
not disagreeing your point. I fully agree where the metal hits the ground, your point is valid as that's what ultimately counts for art and ethics around the topic. I'm only pointing out that the folks voting and making decisions on the space... have a different optic on the situation (one that's even worse in my opinion)
Thank you for your videos Matteo, I love your stuff. 🙌
Thanks bro for such deep video! I agree with your position and believe that we should focus on mindfulness, ethical usage of AI and promote it. It can be very helpful in the right hands. Again, Matteo, Thank you ❤🔥
This was a thought-provoking and well done video. I appreciate you taking the time to step back and think about the bigger picture; and of course I'm deeply thankful for all your work to create and teach SD tools. A few additional thoughts (please forgive the length, but I wanted to be as rigorous and nuanced as possible):
I agree that some people will always complain about models that are anything but 100% free and that these complaints are often misplaced. However, I think the bigger issue with Mobius is that it was not clear what that model was actually offering; and, when faced with questions, the creators weren't able to give good answers and instead became outright hostile. Then there was subsequent evidence that it might have actually have been more of a scheme to boost a crypto project than any sort of actual breakthrough when it comes to new and useful models.
As far as energy use, it's important to first keep in mind that these are just projections. But even if we assume they turn out to be correct, we need to apply some perspective. 4% of global electricity use equates to (at most) about 1% of global GHG emissions. Not nothing, but still small compared to things like driving, flying, and even air conditioning. If people are inclined to focus on the small contribution of AI before tackling these other bigger things, their issue is more with AI than with climate change. Another important consideration is the fact that data centers are often colocated with renewable generation, so associated emissions are probably lower for this particular use. All this said, the solution remains the same: more renewable electricity generation so that we decarbonize electricity, no matter what we're using it for.
I could say even more, but just want to finally note that the trash "island" is the Pacific is a common misunderstanding of something very real. It's true there is an area of concentrated plastic waste in the Pacific, but contrary to popular belief, it's not an island, most of it is microplastics, most of what is not microplastics is fishing nets and fishing gear, and about 20% of it (as of 2018) was leftover from the Japanese tsunami. It's truly mindblowing to read about: education.nationalgeographic.org/resource/great-pacific-garbage-patch-isnt-what-you-think/
I agree on everything you said.
Regarding the "island" I was reading the article as it was written. It's actually a "patch" and it's estimated to contain 100,000 metric tons of garbage (the article you posted is unfortunately outdated). "island trash" is more sensationalist but it doesn't mean that it's not there. I would also like to point out that the one in Hawaii is just one the 5 patches of garbage in the oceans.
Thanks for all you are doing Mattéo. 😇
If I may share my opinion : I think Art is in the eye of the beholder. So If someone sees Art in your creations, then you're an artist, whatever tools you are using.
Contemporary art teaches us to free ourselves from technique.
defining art is not easy and I don't pretend to know the answer.
generally speaking art requires "intention", it has to have a cultural or historical relevance, it needs to transmit a message... and even the lack of message is a message.
Exercising visual arts doesn't mean that you are making Art. Something can be beautiful for you and you can tell yourself that it's art and that would be 100% true for you, but if you show it to some friends and they say it's not art for them. So what is it? Their vision is as good as yours. If we are talking about Art (capital) we need some kind of importance for the betterment of humanity
This is a great video Matteo. I have a different point of view when you refer to art, but it's refreshing seeing opinionated discussion on this topic that isn't charged or evangelizing the viewer. I feel like much of the way creators can monetize their work through platforms such as youtube, patreon etc incentivize tribalism and hinder collaboration if the creators want to get the most out of their effort.
It would be great having an option such as foldingathome, where individuals can put their hardware towards community-driven collaborative training efforts. I think that might happen in the distant-ish future and will be a significant advantage for local models, allowing them to be trained without relying on companies that might drop open source at any moment. Hopefully no future legislation gets in the way.
Amazing video. I put off watching it because the "Is AI art?" topic usually devolves into a misunderstanding of how AI works from traditional artists and disrespect or disregard of the vast history and importance traditional artists have contributed to culture from the AI users. I appreciate seeing an argument being made that has respect for both sides with an understanding of each perspective. This is an insightful video, thank you for making it.
Well, it is not easy to find someone you totally agree with everything they say, and you're not an exception.
Don't sell yourself short. Your dedication and efforts make you one of the most important individuals in the world of AI. That was my opinion, and today, with your insight and talk about ethics and controversies of the use of AI and its future, I am more convinced.
Я живу в самой зеленой стране на свете. Где 50% вырабатываемой энергии уже зеленая(атомная и гидро-энергетика), а остальные потребители находятся так далеко от центральной части моей страны, что огромная доля энергии попросту расходуется в ничто.
Так что да. Я могу генерировать сколько угодно вайфу. Не беспокойся, Матео. Ты делаешь крутые вещи!
the problem is the training not much the single generation, the waifu thing was a joke (of course)
I have said it many times before and I'm going to keep saying it until everyone is saying it. ALL models should be open source. It should be the law that they have to be. They are created from us they would not exisist without us they should belong to all of us. *We need to fight for this.* Otherwise what he says in this video will be true. All AI will end up belonging to large corperations which is exactly what they want. Corpos don't want us to have this power without paying them for it, even though its build off the backs of every person on the planet.
unfortunately that is the most likely scenario. Adobe, Meta, Google, Apple, Amazon, ... they will own most of it
Thanks Matteo for your videos and thoughts. Those of us who have been around for a few decades recognise the repeated pattern of condemning new tools as they are perceived as threats to artistic integrity. As you point out Photography is an obvious one. I lived through the introduction of personal computers and the condemnation from artists the moment people started using them to produce images. Now of course a vast number of artists use computers exclusively to produce their art. AI or the algorithms behind AI are of course more nuanced and complex but I agree are still a 'human made tool' that can be used to create artistic expressions. It nearly always comes back to who is behind the tool and what is their intent. Already I see creative people making wonderful images that stand out from the templated button pushers. The fusion of traditional, digital and AI tools is definitely one way forward. Just looking at art history I have little doubt a new expression of art will be born where the use of AI is central and eventually accepted by the majority. It might just take a few decades...🙂
I think you are wrong on this one.
What is art?
Art is what conveys feelings.
What is Art is decided by the audience, not by the artist.
If you see a scene of a homeless guy with his cart under the rain, you feel something, yet there is no director behind it.
So, can generative AI produce art?
Of course, it's all in the audience reaction, in how an image clicks with the watcher.
Now, the real question is, is the guy who simply types prompt an artist?
No, the artist is the AI, the guy typing prompt can call himself an art director, which is the one ask the artist what to do, and makes a selections on what the artist produce.
Great video on an important topic.
Thanks for comments. I took a similar path as you. I was a fine arts major but worked in IT as a career for 30 years because it came natural to me. I still draw and paint and visual artists that make their living this way are right to be concerned. But I also think people who know and enjoy art will always appreciate and want the hand of the artist. Also, AI cannot sculpt yet 😂. Michelangelo felt sculpting was superior to drawing and painting. Just another point of view. Thanks for your great work. I hope you keep going helping us tinkerers.
Thank you Matteo.
I share a quite similar background with you, I have a Bachelor degree in photography and I'm a self taught designer, as well as coder. I find it fascinating how AI allows me to connect art and tech in a really novel way.
I also share the same feeling that AI is a tool, that can be misused, but can also be used to create wonderful pieces. I'd even go further and say that creating good tools (like the ones you create) is also a form of art, just like we value a good paintbrush made of good materials and we pay more for it.
Thank you for posting a wonderful topic. I really hope that the tools we have today aren't placed behind lock and key. I'll be honest I love art but i am not talented. These tools have granted me the ability to express myself.
It was a great education and a great inspiration and lesson.
I did not expect this man, I completely agree.
that was refreshing & I'm glad you didn't give the snake oil tech-bro answer i came to expect from AI channels
I was astonishing when I Saw the picture of Baudelaire
I am a trainer and the moment training can't be done locally, for whatever reason, is the moment I uninstall all this. Having said this, and as said trainer, I am sick and tired of nothing being done for the advancement of training. I see some advancements for inference, but we are using the same tools for training I used with 1.4 (under the hood) yet not even the models are the same. Basically, I am trying to write my War and Peace using stone tablets and a few chisels. Not sure why there have been no real advancements for us in training.
Is this referring to sd_scripts? What kinds of improved training tools do you have in mind?
@@shisofox All of them, but if you look they are all doing about the same things but no advancements. In inference we have advancements, and stuff we never had in 1.4, but why am I stuck with the same tools, and pipelines, since then for training? I'll tell you why because the companies are buying up the talent that knows how stuff works and could get it done. Most of the people I hung out with, or around, are now working for SAI, for instance, or other companies, so their githubs are either stagnant or shut down.
@@generalawareness101 I'm not nearly as connected as you. The only thing I remember learning about is vision models for generating better captions. It sounds very expensive. Can you ask your former contacts what tools they have that might improve training, that don't cost millions to run?
@@shisofox There is a difference as they are in a business while I am talking about the advancements we have seen in just inference. 1 step gens, 8 step, etc... those are peanuts. Look at IPA for inference. That is a huge advancement for inference. Look at control net, etc... these are all advancements for inference, but we are still using the same training tools we have been. Over on the LLM side people have told me the tools they have they wish they had on SD as they would be phenomenal, but I can only take their word for it as I don't do LLMs.
My whole point is the training side is just flat out stagnant, while the inference side has some movement.
Bravo Matteo! 🎉🎉🎉🎉🎉🎉
Keep up with the good work
Thanks so much your video has given me enough confidence now to publish my ai genersted stuff bevauee i will specially convey my viewers that i am using ai art to convey a msg to them .
I really love that you are so considerate and contribute so positively to the community. Do you have a recommendation of how to get involved with projects to start helping the community? Eg. Start by following this git project or hugging face group and go from there?
if you don't have specific technical skill, I guess the best is helping others resolving basic issues. Check the bug trackers, 90% of the times they are very simple support requests.
Hello Matteo،
Your good tutorials changed my perspective on this art, thank you
Please do detailed tutorial on controlnets and frame interpolation
Thank you in advance
Certainly, incorporating AI into the art-making process is akin to using traditional tools like pencils, brushes, and digital software. Just as artists throughout history have utilized various materials to express their creativity, AI serves as a modern tool for artists to explore and innovate. This fusion of technology and artistry gives rise to a new movement, AI-ISM, where the outcome is not just a mere product but a manifestation of human ingenuity amplified by artificial intelligence.
I agree with you! It was not boring and your point makes total sense. I don't know what makes the difference in the mind that makes you want to collaborate with open source and community. I empathize with your disappointment and I don't know what's the solution, I cheers for some of those contributors become a big company and payback all the unbalance between those who uses and monetize and those who gives something back. Good video
Im agree with you,.. specially the word " AI is just a TOOLS "
How to download the workflow at 11:26 in the video? Thank you.🥰😝😀🤩😍
I agree with a lot of what you have said. To me ai is meant to be used as a tool and those of us who learn it's intricacies will be propelled forward. Of course there will be those who will abuse it in undesirable ways, but that can also be said about most anything we already have. In time tech will catch up so we can much easier filter out and find all the ai generated stuff so people don't get fooled when it's used in an exploitative manner (politics etc). Right now it's still the wild wild west in terms of control/regulation/law so of course many ppl will get fooled but i think that in itself is something we have to go through to develop a safer future with a i.
It's too much a useful tool to put back in the box, and so much more can be done with it in many industries.
Well its true, we have increadible problems regarding our lack of sustainability. This was a jurney, i started 10 years ago as an awakening. The real Problem is, not climatechange for example, all these problems ar connected to another and not separated Crisises. They are conected through only one single metacrysis. Called Overshoot. So at the moment we use up even our renewable ressources about 70% faster as they can regenerate. like a bankaccount, you could live from for ever if you just used the interest. But that's not what we are doing. We are taking the interest of about 2.7% of the biological capacity of the planet, plus another 1.5%, by eating up the deposit capital, so to speak. The consequences: overfishing, loss of forests, topsoil, biodiversity, etc.
And we are not yet talking about non-renewable resources. Especially when it comes to energy and energy resources, we are living in an imaginary fairyland. The only reason oil has not peaked is because fracking in the USA prevented production from peaking for 15 years. But this will not happen again. Alternatives are not in sight fast enough.
It will be a bumpy ride into adulthood for humanity. With the realization that limits are real and trees don't grow to the sky. And AI won't change that.
For those of you who would like learn about these topics from one of our greatest minds, Prof. William Rees, who was one creator of the ökological footprint
th-cam.com/video/LQTuDttP2Yg/w-d-xo.html
And for this channel a talk about Ai and Sustainability from Nathe Hagens and Daniel Schmachtenberger.. very deep, but excessive long, but worth it..
th-cam.com/video/_P8PLHvZygo/w-d-xo.html
(one impressive center statement from the whole interview above. if 3 hour are to long... just start here. but its worth the effort as well as the ökological superhero william rees in the first link th-cam.com/video/_P8PLHvZygo/w-d-xo.htmlsi=Sn9LnmTbwHDAk0VB&t=9946).
good talk!
Hi Matteo, thanks for your work!!! May i introduce Nicéphore Niépce, He is the inventor of photography in 1827, then called “heliographic process” the invention of a new industrial print "multiple": this makes sense with the evolution of computer neural network technology. Maybe not yet an art
Inspiring ❤
As a traditional artist, I am not so concerned with the philosophical questions of whether AI art is art (although I don't think that discussion is invalid), but rather with the virulent resistance of some institutionalized artists to this powerful tool and its possibilities. Beyond the ideological biases in this resistance, there is also a lot of snobbery, a belief in being very special individuals. It would be unfortunate if, trying to prevent their style or work from being "stolen," they instigate the imposition of laws that aim to limit the use of generative AIs, but end up favoring large tech companies more than protecting artists' turf.
Agreed, although I'd still feel much better about using these tools, if living artists who didn't give their concent would be excluded from the training data and those who did, would get some compansation. I avoid deliberately prompting for living artists for everything I share publicly, but that approach of course only gets you so far.
Attention AI companies: Hire this man to guide your model development, sit on the ethics board, and go on TV as the face of your tech. And Pay him a lot to just code!
LOL, thanks for the morning laugh 🤣
I think the key element of art makes you feel something is that it must be an intentional expression.
In this way an autogenerated piece of imagery doesn't necessarily count as art even if it stirs and inspired the viewer. If however someone chose the location to present the autogenerated piece and timing to show it, then it becomes part of a performance art piece. The key is people intend to express something, which is the artist feeling something and using the art medium of choice to do so.
You forget that AI only derives its creations from already intentional human art. Even a drawing style has an original intention (ex, cubism) and is called an artistic style, which is encompassed within the representations of human art.
And this would count even ignoring the fact that the person who asks the AI can provide the intention.
And that ignores the fact that there are infinite ways to create with AI beyond just writing an order and that's it.
@@cesar4729 i didn't forget at all. i simply stuck to the point i was trying to make.
My perspective is that what a person makes with their chosen medium, regardless of the tools used, is an intention to express something they feel which is what makes it art.
@@magejoshplays That is too rigid as a definition. It means that if I make a sculpture for no reason, no one has the right to find artistic value in it. Intention cannot be measured, nor can an author from other centuries say if he did something with any intention.
The world does not work that way. It is the observers who decide what to call art or not, and it is the majority collective consensus that allows a relative definition of art in each culture and era.
There is *no* concrete definition, since it is a subjective and fluid concept. The closest definition is that art is what people consider art to be in that period.
Say you are an artist and you train a model over your art. Then you use that model to speed up or augment your work, say to make backgrounds. Then you improve the generations by retouching them, adding elements, fixing colors. AI becomes part of a process and in that case in my opinion completely transparent.
That in my book is a good use of the technology.
Art can be many things but I mostly agree that needs to be intentional. Even though it's not always the case. It can be defined art by "history" or for its cultural relevance. Think for example of the "Moka pot", it was not intended as art but it's now showcased at the Moma
@@cesar4729 I'd suggest that even if you label it as no reason, you still in fact had reason and it therefore becomes art.
Then I'd suggest that if someone else comes along and decides they interpret meaning from that expression, they have also defined it as art even if not consciously.
Because, as you suggested, ultimately all meaning and definition is entirely relative to the perceiver and all arguments otherwise are simply a continuation of the argument of sovereign soul or subject to a master.
Congrats ❤💯
What are the server side issues with ComfyUI
we should discuss more about ai and jobs
I worry about food, drinking water and breathing air,thank you!
Very interesting and thoughtful. As a professional painter and illustrator AI has been a fascinating journey for nearly a year. I still don't know what my feelings are. Maybe one indicator is that I would never sign a Comfy generation. I don't feel that I made them. Perhaps all images are somewhere in the code and we are explorers seeking them, rather than artists creating them. My painter friends see it as a threat, but no one will want an AI image framed on the wall, sure commercial artists are going to feel a great deal of pain, but people will still want hand created items in their homes. I do wonder what worth the tsunami of images has, no-one gives most of them more than the briefest glance. Is their value in the brief moment of pleasure the creator feels? Maybe, there seems to be more interest in workflows than the end result. When I post videos I only get comments about how they were made and almost never about the images themselves.
Yeah, so far it feels a lot like building sand castles, being proud of them for a little while and then have them washed away by the waves the next day. I'm usually more proud of my workflows than of the images they produce. But as soon as I find a real professional usecase for all of this, maybe that feeling will change.
In my opinion, Art is not the media, the Technik or the material used to create something. Of course mastery used crafrmansship for what ever shows the love someone puts in his efforts and this creates things of wealth and beauty on its own. But areal piece of art for me is something else. It needs to describe a thing a feeling a thought in a way that was never told before. It should enrich our point of views about this world to get a bigger view of reality or to say enlarges also the reality of mind.
We really need these additional tools. Language is fine for AI dialogue generation, but it is too scarce for images. In reality, it is difficult for us to use language to carefully annotate a picture, and then apply this annotation method to AI painting. It is difficult or impossible to achieve the picture you imagine simply through prompt words. You can only use Lora or ControlNet to tell AI the desired picture effect through picture to picture. That means you must have pictures first to generate the imagination in your mind. It is a very painful logic. Will there be something beyond prompt words in the future?Connect my brain to controlnet!
Thank you Matteo for your hard work! In the first part of the video, you mentioned competition and what impetus it gave to the development of art. Unfortunately, generative technologies have been taken out of the competitive field. There is still no clear idea who is the author of the generated image. Is it possible to use it for commercial purposes? There is also no legal assessment given for collecting material for training models. All this rather slows down the development of art.
Thanks!
Imagine all AI companies merging their LLMs and Generative art models, we are affraid that AI will outgrow us but as long as these companies keep competing with eachother training their own models with 70 billion and 400 billion and what not parameters, think of what a merge between Llama3 and OpenAI would cause, that merger would blow EVERY other existing model out of the sky and AGI would probably arrive a lot faster than expected, but as long as they keep competing AI will save no one and the energy consumption will grow as is
thanks :)
Preach brother.
I only disagree ( and not completely) on prompt engineering. Yes it does not make you an artist, as much as knowing how to get the best out of a DSLR camera does not make you an artist, but it sure helps out a lot.
And it makes you more proficient and capable of getting art out of your tool. It's a skill that contributes to your art.
I agree on everything else.
If, as a community, we could rebuild insightface training datasets and re-train on it to release open source weights it would be an achievement.
Thanks for all your work
I'm not saying that "prompt engineering" is not important. I'm saying that it's not engineering and doesn't make you an artist.
Also it's a term invented by the AI companies so they can justify the loss of jobs over AI... because they give work to "prompt engineers"
@@latentvision I'm not entirely sure about it's definition. I believe it's real application is in alignment and model weight "exploration". It is used to try and understand the model propagation, limits, and strong points. And sure enough is not how the term is used normally, we might be more in the field of prompt design. It's a big AI company term because only them can really put engineering behind datasets. Yet lllyasviel recent Omost still uses a form of prompt engineering and canvas control to achieve quite interesting results.
I feel like I'm debating on terminology just for the sake of it, I'm sorry 😅
LEGEND
Sooner rather than later, AI will be so deeply integrated into countless areas that all the current complaints against AI-generated art will age poorly.
This is the same for photography or films its new so of course people will look down at it at first.
Is painting by numbers art?
to be honest... anything can be art
🙏
Artists will call a red sqare painted over a green background Art, and then freak out at AI created images.
Jokes asides, I find the "Is this art" debate very boring. I find the AI negationists dull and annoying. If AI is a tool, whatever you produce with it could be art. Same goes for a brush, photoshop, a guitar or a mixing board.
Quoting Robert Fripp: "I don't like that word, Art. If one sees art as that breeze that bloweth as it listeth that intangible quality, that certain something, the magic of a performance... we can't control that. But we can raise a sail to catch the breeze."
And here we are, debating about the cloth of the sail.
Hey we have the same work backgorund! :)
First of all many thanks for all the work done and for the tutorials! Then i guess that gen ai is a bit like ebikes: if I use one I can make 50 km in a day, if you give it to a pro they can make 500 km in a day. So, I would say, if you want to make art with ai you'd better know some colour theory and composition, or if you want to do brand communication you'd better know some commercial art direction. The same applies to language models. Last but not least: are plastic testicles good with sex dolls? 🤔
don't know about sex dolls, but they are perfect for waifu pillows! 😆
@@latentvision 🤣
A scribble by a 3-year-old is still art. It's a unique product that was created over time by applying skill and imagination. It might not be art that will impress many people in terms of technicality or innovation, but that's not obligatory.
There is Art and there is "practicing visual arts". We are most likely all doing the latter.