Free AI Image Generation: Demos & Dangers

Free AI image generators Stable Diffusion, Bing Image Creator and Leonardo AI demonstrated, followed by a discussion of the broader implications of this new family of AI technology.
Links to all of the applications demonstrated or otherwise mentioned in the video are as follows:
Stable Diffusion Online: stablediffusionweb.com -- note that since the video was recorded, this now requires the creation of a free account, or to login using Google.
Bing Image Creator: www.bing.com/images/create
Leonardo AI: leonardo.ai
OTHER AI IMAGE GENERATORS THAT CAN BE USED FOR FREE:
Nightcafe: creator.nightcafe.studio
Craiyon: www.craiyon.com
Lexica: lexica.art/aperture
Playground AI: playgroundai.com
Gencraft: gencraft.com/generate
Meta one: imagine.meta.com (not all regions)
PAID SERVICES:
Midjourney: www.midjourney.com/ or www.mymidjourney.ai/
Google’s Imagen-2 deepmind.google/technologies/...
REFERENCES FOR FINAL SECTION:
stablediffusionweb.com/#faq
www.theverge.com/2023/2/6/235...
www.theverge.com/2023/12/4/23...
openai.com/dall-e-3 -- look to opt-out bit, the open form and highlight cat bit! :)
For additional ExplainingComputers videos and weekly updates, you can learn about becoming a channel member here: / @explainingcomputers
More videos on computing and related topics can be found at:
/ @explainingcomputers
You may also like my ExplainingTheFuture channel at: / @explainingthefuture
Chapters:
00:00 Introduction
00:48 Stable Diffusion
04:21 Bing Image Creator
06:59 Leonardo AI
11:55 The Potential Downside
#AI #StableDiffusion #BingImageCreator #LeonardoAI #ExplainingComputers

[@ExplainingComputers]

Sadly, as many have noted here, the free version of Stable Diffusion now requires an account to be created, or to log in using a Google account (although use is still free). It worked without a login when the video was made in mid January 2024, and indeed still worked for me without a login on February 4th 2024 when this video was released. But by February 5th, I found that a login was required. Perhaps this was a caching issue.

[@TheAnkMan]

Yeah, suppose it was your video. They must have received tons of hits. DeepAI seems still to work without registration.

[@gr-os4gd]

You can still self-host (it's easy) w/o an account.

[@DavidFrankland]

Overwhelmed by viewers of this channel

[@sirant]

It also has a credit system, limited to 7 initial credits. BTW, your Telegram scammer is still here sending messages. Beware everyone!

[@d.barnette2687]

@@sirant Thanks for the heads-up. I just received a message from the Telegram scammer and was wondering what the heck it was. My comment said there was a reply, but there was no content. However, my email had the scam message that I needed to reply to @explainingcomputers. It all looks legit, but I thought why would I get such an email except as a scan. So off to the trash bin that email goes.

[@generalawareness101]

What gets me about Getty is that a lot of the public domain images are now only through them that I have to pay to see without their watermark. I remember when certain images were freely available on the net, then I had to find one I needed again, and it was suddenly behind a pay wall. These are historical images they do not belong to Getty yet the only place I could find some were on them.

[@einglis]

"There are various kinds of cyborg panda, as you may be aware". Gold.

[@BrockMcLellan]

In an obscure part of rural Norway, some of us decided we needed a change from conventional painting with brushes, and have decided to create artworks doing something different. Yes, we have acquired airbrushes and compressors!

[@urbanimage]

Steady on there.

[@ridefast0]

Getty Images gets no sympathy from me - in the early 2000s they acted (shamefully IMHO) to instantly fine small and residential/family website owners after their web builders made errors in image selection, with no option to cease and desist going straight for a legal-threat-backed money grab from the end user. Yes I was fined and it still makes me angry!!👿😠

[@Praxibetel-Ix]

Oh, wow, that's just really crooked of them. I'm so sorry you went through that!

[@ridefast0]

@@Praxibetel-Ix Thanks, and I think there were many volunteer and charity organizations hit worse than I was! It turns out that web design contracts in those days often left the end user in hot water rather than transfer the risk to the web designer who misused the images! (End of rant).

[@minotaurbison]

Before I type this, I want to clarify, this is a compliment and not sarcasm or any sort of insult. This guy reminds me of 1980's PBS education TV, very easy to watch, to the point and always keeps my attention. Great job! I was a child in the early 80's and PBS was my favorite channel back then... we didn't have cable and only had 4 channels total so having one that I liked was great. Watching your videos brings back fond memories.

[@ShizukanaEntertainment]

As an AI myself, who've been watching this channel for some years, I was shocked to see no ducks or scissors in the prompts.

[@FranklyPeetoons]

Excellent basic introduction. SDXL, with light tinkering, can be made much closer in terms of quality to DALL-E 3. Additionally SDXL can be self-hosted - for free - if a suitable GPU is present. A number of interfaces are available, and the customization options are truly staggering. The best compromise between SDXL configurability and advanced features may be FooocusMRE, which runs in a browser window. Advanced features can be toggled on or off. A bonus of self-hosting SDXL is that all content restriction is stripped out. However, I can confirm that the middle- to low-end illustration market is well along the way to being wiped out by AI. Blogs, news services, and advertisers who use AI art often don't even spend the meager time needed to correct AI's distinctive & frequent visual imperfections.

[@DavidDoyleOutdoors]

I also use Fooocus, haven't tried any of the forks yet such as MRE. I personally think SDXL is superior, it produces more photorealistic images, whereas DALL-E 3 is more cartoony

[@stabokbose]

Sunday+ Explaining Computer video >>>>

[@d.barnette2687]

Greetings from across the pond near Albuquerque, New Mexico, USA. Great AI Image Generation video. Now every time I have a bug in my Python program, I will do my best to resist the urge to look inside the computer to see if there's a pink spider crawling over my processor. Well, at least we don't have vacuum tubes anymore. As you know, back in the day real bugs would crawl inside a computer and short the tubes, causing a running program to fail. The whole concept of having a programming 'bug' today comes from this. Nice to think the bugs now have some color to them, I reckon.

[@ExplainingComputers]

Thanks for your support, most appreciated. And do look out for those pink spiders.

[@MikeBob2023]

*Thank you* for your thoughtful dissertation on the implications of AI being used in this way. I had no idea these image creation sites even existed, much less that their use was quite so widespread. On the one hand, the ability to create such images via AI seems novel and fun; a 'toy' for creatively challenged people like me to play with. But I also see it as a veritable 'Pandora's Box' in terms of its potential impacts on those whose livelihoods are based (and dependent) upon their own 'human' ability to create such imagery. I believe that in your part of the world this is known as a 'sticky wicket.' (Thanks again. 👍🏼)

[@maxdiamond55]

Thanks Chris for a fabulously insightful glimpse into the world of ai image generation. As you infer, this rabbit hole continues to deepen, hopefully not to the determent of future artistic creativity.

[@SuprousOxide]

Love that double ended utensil the bing rabbit is using. Knife on one side, fork on the other. Like some crazy D & D weapon...

[@Praxibetel-Ix]

It's Stanley the Knife's cousin!

[@CCoburn3]

Because AI doesn’t understand anything, it makes mistakes that children wouldn’t make. Even a 2-year-old knows that the utensil wouldn’t work. But AI doesn’t understand anything. So it creates some really whacky images. (By the way, it is well-known in D&D circles that the proper way to defeat a rabbit wielding a knork is to use the Holy Hand Grenade of Antioch.)

[@alttiakujarvi]

Bing clearly did not understand, what eating with a fork and a knife meant, as none of the rabbits were actually using the fork and the knife for eating. They just had then *with* them.

[@bobrowles43]

Thank you.. Great content and thoughtful discussion on a very complex issue.

[@KISSbestfan]

Dear Chris. To be absolutely honest, AI was a huge concern of mine in the following year. Everyday I See more and more businesses clearly using the AI generated images, which you can still easily spot with a bare eye, as those always do have imperfections. I checked over today at KZbin, and I See your (Another) video on the subject, So I Just had to watch it and write a comment. I know it has been a very long time since I was active over here. To be fair I do not think that our last job would be connected with AI, as dozens of economy branches had relied, and would still rely on the manpower. As for now its purposes are So far only limited to small stuff, So Im waiting for the first huge time one, like creating a commercial using fully AI. Maybe I wouldnt be So concerned, if not the Fact that a lot has changed in the society, and for many Young people, AI May become a part of their Everyday life, instead of being hungry for True experiences. Blimey, I forgot about the roast in the oven. This is how addicting the subject can be.

[@Praxibetel-Ix]

Don't let that roast burn!

[@Darkk6969]

@@Praxibetel-Ix Need to get one of those AI controlled oven!

[@Praxibetel-Ix]

@@Darkk6969 I can already imagine the fires!

[@Reziac]

Commercial and cover artists are now using generative images as samples for clients, to save time and money -- they don't have to hand-crank the art samples for the client. Then once the client picks what they want, the artist spends his time and the client's money creating the real thing, with more likelihood of satisfaction on the first attempt. There are already cases where AI coughed up something the artist didn't think of. In one case (I've seen the result) apparently the prompt was effectively something like "desperate chained slave girl" ... and all by itself, the AI added a mirror containing an image of the girl's evil twin, and perfectly composed the layout to best effect. Quite remarkable.

[@ukar69]

I used Bing in helping create specific images for someone's KZbin thumbnails. A couple of requests I made came out OK but no matter how specific I was, it just couldn't come up with an empty car park at midday. Every image had some cars in it and the lighting clearly wasn't midday. There's still a role for human photography. Just.

[@generalawareness101]

I am an AI model.lora/loha trainer and I see that I would immediately grab those images and train it better. The problem with all of these, that Chris showed, is that they are base models and with users who can train on them have they blow away the for pay stuff. This is why the web based only approach is the worst as it is static waiting on their company people to make new models, and new low rank model for them.

[@Mrvideosandgames]

Just tried with "deserted parking lot, midday" and two of the four have no cars. But the lines it painted on the roads are more than a little wrong. It's unclear what time of day the pictures represent, but judging from the angle of the shadows it isn't high noon.

[@alttiakujarvi]

there is an old joke about a fisherman. A neighbourgh asks him: - What are you doing? - I'm fishing pikes. - Have you caught any? - None. - Well how do you know you are fishing pikes? It seems that Bing does not no if a car park is a car park, if there are no cars there.

[@perrymcclusky4695]

Definitely a fun, interesting and thoughtful video. Looking forward to your next video!

[@alanthornton3530]

I was amazed with Leonardo AI, I thought the image quality of the Cyborg Panda & Spotted Rabbit was impressive, the lighting was spot on. In regards to intellectual property & copyright you've raised some excellent points! As an amateur photographer I upload my images to Flickr, I own the copyright with 'All rights reserved', I'm also a member of a Photo group a lot of our competition entries (photos) are packaged into a slide show & uploaded to KZbin & Facebook, I did raise the question of permissions/copyright no answer came the stern reply!! Art students copy the Old Masters as part of their studies, if they then display their work it's attribution to the original artist is noted. Thanks Chris for a thoroughly enjoyable video, kind regards Alan :)

[@ExplainingComputers]

Thanks for your support Alan. :)

[@s0men00bb]

The same is in technical field. Like , when we hear , very recently - Tesla invented this or that , that is only partially true , as , with everything else , you need base for something , nothing comes out from thin air. :) So , if you wanted to use material I made during my years in school and university , you'd have to quote me as source and resource used , and this is how is ball kept rolling. :) What struck me the most , as I'm terrible in hand drawing but excellent in geometry and technical drawing is - there are the books that teach exactly - that , using geometry shapes and lines (that get removed or overdrawn or painted over the lines that were used as a base) to achieve great results. To me , it was like old dream coming true , as I can finally use that to create something and not be stranded because of lack of talent and skill and turn it into something ! 😁

[@Okurka.]

Leonardo AI created a panda with 6 fingers on one paw.

[@ahmad-murery]

This is really impressive, Use AI as a tool, otherwise it may enslave you in a way or another. Thanks Chris!

[@arvaneret_329]

Your comment needs more likes.

[@Druac]

That is all AI is…a powerful tool.

[@adamo1139]

I feel like a video about Stable Diffusion is not complete without telling people that model can run on your own PC with no internet required and you just pay for electricity. You can run Stable Diffusion, Stable Diffusion XL and Stable Video Diffusion all on your own PC, best if you have somewhat recent GPU though, but it works on CPU only too.

[@DavidDoyleOutdoors]

I ran SDXL on a GTX 1080 6GB, it was slow, about 3 mins per image, have now upgraded to a RTX 3060 12GB and now its only 30-40 seconds per image

[@Druac]

Exactly…building a home server at end of year to do just that…but hoping the 5090 is out by then *crosses fingers*

[@Naa-ee7nq]

those are some stellar prompts, you have a future in vibe engineering

[@williamguru]

Artists are going to create no matter what. If anything, AI will open new avenues of creativity.

[@thomasgraf1986]

Wow, fantastic video Chris! A really great dive into the amazing potential and negative drawbacks of the AI future.

[@chromerims]

Our future selves will be so thankful for (and delighted by) EC's chronicling and analysis of technology's progressions during these times 👍 With contemporary fine arts, an unspoken and compulsory routine abounds for artists to author a concise but deeply insightful verbal accompaniment with each of their visual or sonic oeuvres, whether such words be for explication or more simply contextualisation. Seems an analogue a bit to the AI tags and prompts. Whereas such routine had manifested less frequently in the past with respect to decorative arts, which works more often than fine arts stood apart from text descriptors. Kindest regards, neighbours and friends.

[@JohnDunne001]

Another great video - thank you for all your effort! I think this recent wave of AI is about non technical types having the tools made available to them to leverage AI. We're seeing businesses adapting to this wave of adoption but as always, we'll figure out a way to put this to good use and inevitably that means some people lose out. It's a shame but that's the nature of technological progress. It's been the same since the screw/steam engine/IC were invented.

[@cheesyboofs]

Really great video Chris! I use Stable Diffusion locally and I enjoy coming up with the most imaginative tailored prompts I can to generate images I could never 'create' with a paintbrush and paper. I suppose like all technological advances we might as well embrace it as its not going anywhere any time soon and if anything it makes 'real art' even more valuable?

[@zzstoner]

Sometimes the UI of some websites can be confusing at best, or just plain wrong at worst. Today's case-in-point is for Stable Diffusion's site. As he enters the prompts, the main *"Prompt"* (at 1:35) has the grey "helper text" of _"Enter your _*_negative_*_ prompt here"_ ... ...whereas in the Advanced Options, the *"Negative Prompt"* (at 2:05) also has the same helper text of _"Enter your _*_negative_*_ prompt here"_ . Naturally, the helper text for 2nd prompt intuitively makes sense for its intended purpose, while the 1st prompt has the wrong help text by suggesting to do the _opposite_ of its intention. Moral of the story to Web Devs out there: CSS is not always the holy grail solution for styling/layout, as it will never catch the human intent of each UI control or element. 😸 👮

[@computertuts623]

Thanks for this. I've been wanting to get some images generated for quite some time, but it seemed the barrier to entry was a bit high, but you made it look obtainable.

[@mikefinn2101]

Peter thanks for giving me a great introduction into AI, glad you understand this and have taken the time to do this. an Eye opener for me. It is the great kindness of people like you that help others like me keep up. Grateful to you for all the amazing help you provide me. All new to me.

[@acerforge3882]

A thoughtful and interesting video. Aside from all the other very valid comments I also question the terminology when there is no obvious intelligence ( in the human sense) displayed by the programs themselves. For example almost any adult on the planet would know that virtually all spiders have 8 legs but the program does not apply that constraint because it was not told to. Much the same could be said of those gifted individuals who can faultlessly recall facts about any topic they have encountered in their entire lives. Gifted founts of knowledge but not necessarily intelligent!

[@ruudb72]

Excellent video and nice examples setup!

[@MrKelaher]

I have met the Leonardo peeps. Nice folks ! Re the copyright issues - the models are mathematically perhaps best understood as very high dimensional holograms containing many intersecting compressed images, the intersections being their commonalities from an affine transformation perspective. The original images used for training can in fact be largely recovered if a blurred corpus image is used before iteration commences.

[@gfdggdfgdgf]

Local ai generation is the way to go

[@terrydaktyllus1320]

For what purpose, precisely? I see the benefits of AI coming out of scientific and medical research where there is a need to do "intelligent" analysis of huge amounts of data that may reveal patterns and trends that lead to great discoveries in the future. All that stuffed is "championed" by big corporations and governments, not the likes of you and I as private individuals. And why do I need AI locally if I'm already used to thinking for myself anyway? I write a lot of short BASH and Python scripts for my day job and my home computing environment and I know ChatGPT can write me a working script in seconds because I tried it once. But the "fun" for me is doing the writing with the satisfaction of "a job well done" when it works. Having the ability to generate these kinds of images locally would let me create some interesting wallpapers for my computers occasionally - big deal. As far as I am concerned, AI goes hand-in-hand with the lowering of standards in education and critical thinking not being taught to younger people in the "me me me, now now now" generations as they turn their brains off.

[@toucan221]

@@terrydaktyllus1320 Very well said, Thanks

[@terrydaktyllus1320]

@@toucan221 Thanks, I appreciate your compliment. I simple believe in keeping my brain active so that I always have the choice to use AI or do it myself, rather than just turning my brain off and giving into AI fully.

[@anonylesss]

i don't have an opinion on using AI for image generation. but, I have tried ChatGPT and Bard for troubleshooting Linux issues. they both gave me wrong answers but with great confidence and detail. this is to a point where a novice Linux user would have make his PC unbootable

[@terrydaktyllus1320]

To give ChatGPT some credit, it wrote me a BASH script based on a description that worked first time when I tried it - but, the satisfaction of achievement was not there and I still write scripts myself for that reason alone.

[@brianwood5220]

Very thought provoking. Thanks for sharing, Chris.

[@TheSillyshyguy]

I really enjoyed this video. Many thanks, Mr Chris :)

[@benarcher372]

Great video. Fascinating subject.

[@LMacNeill]

I think a part of the solution to preventing AI from taking over all creative endeavours is to ensure that AI-generated images, music, text, etc. cannot be copyrighted.

[@MicrobyteAlan]

Great episode, this will help me create imaginative 3D prints. Thanks

[@robertlawrence9000]

Nice comparison video! One thing that makes me think is how possibly distorted AI generation of art might be later when enough AI images also flood online where the AI art generators are learning from. Maybe one day a "Cyborg Panda" will look less recognizable as it's learning from "Cyborg Panda" images online and thus distorting the next image outcome even more.

[@ragingbombast]

I read an article (That I cannot for the life of me find now) that investigated this. When AI art generators are trained on AI art, they basically start to crash and produce gibberish. The same thing happens to things like ChatGPT when it starts feeding on its own output. I suppose eventually the people designing these systems will have to find a way to filter out AI generated content from learning modules, and isn't that just delightful.

[@robertlawrence9000]

​@@ragingbombast exactly

[@adamo1139]

@@ragingbombastDoubt it's true. I fine tuned LLMs on their own output, they work fine. It's just text after all.

[@PaulHosler]

You missed trying to generate a Cottage Cheese Cheese Cottage 🤪

[@ExplainingComputers]

Cool prompt!

[@schitlipz]

Ooooo. I just thought of a spooky one to frighten children: Castle made of teeth from tooth fairy.

[@tpobrienjr]

More like Swiss Cheddar?

[@jajwarehouse1]

@@schitlipz Sounds like something from Discworld

[@tonykeltsflorida]

What about a Swiss cheese chalet in the alps?

[@bahshas]

AI is going to completely change society in ways we cant even imagine or much less predict now

[@weirdscix]

I'm a big fan of AI but I don't think they will ever replace artists, I just view it as another tool, I have skills in 3D but absolutely none in traditional art systems, this gives me a chance to be creative for myself. Regarding the training of AI, it should be ethical and I hope that's something which is resolved in the near future.

[@therushden]

thanks mr b, a very interesting insight into ai image generation 👍👍

[@user-jx5wl6sw3d]

My friend is an illustrator of 40 years experience says that it does not matter how you get the image as long as it suits its purpose, how ever i agree with you.

[@toucan221]

sounds like your friend is going put himself out of a job with an attitude like that!

[@MrBradleykeith]

Hi Chris, over the past two years or so I have been invited by people to join a few platforms that will pay you to send in selfies, photos of places and everyday incidents all for this very purpose. We are writing our own resignation letters.

[@tpobrienjr]

In your last remarks, you mentioned the dilemma of the programmer/artist who was faced with training his AI replacement. I had a similar experience in the early days of "outsourcing", when I had to train several people on the support of a large mainframe HR system. I'm recommending this episode to my class (30 senior citizens).

[@bahshas]

at least with ai you can use the tools to your own advantage while with outsourcing youre just being replaced

[@KosmosAc]

AI finally should make us humans question our definition of creativity. In my eyes this definition shifted from "create something new" over to "rearrange what was already there", and this isn't the fault of "creators", it happened on public demand. In consequence we had a massive inflation of people who called themselves "creators" to fullfill this demand. AI is partially able to do this part now. At the same time I think this is a chance for "creators" to really "create" something new again.

[@MaggieKeizai]

Well said. For years I've bemoaned the state of graphic design, where "creative" people just juxtapose mario and a stormtrooper in some kind of way, or otherwise apply some sort of aesthetic to IP that's long since been worn threadbare, and then put it on a t-shirt. Naff, naff, naff. Whatever the eventual fallout, this does promise to be something of a rug-pull for that kind of lazy commercial "art".

[@DaleDix]

I've said it before and been laughed at. People can't add up or use a digital clock. they don't understand grammar and are as dumb as a doornail. The next decade is going to make them so dumb that they will need ai to read it out to them in headphones.

[@akaabbs259]

ecclesiastes 1:9 What has been will be again, what has been done will be done again; there is nothing new under the sun. i would say use both in the future things will become more ez production scales increase we can now focus more on the details and styles leave the rithmic outlines to something else or you can always go back to old fashion ways but in my view art should be used for insparation the cherry on top the cosmetics its not needed its wanted

[@tonysheerness2427]

It is back to the future, isn't this only an update of clip art?

[@CCoburn3]

I agree. AI is a tool. And people who learn to use the tool will be able to create new things. New tools equal new opportunities.

[@Lurch-Bot]

I'm not much of a graphic artist so AI image generation looks like fun. I have the imagination but somehow just can't translate it into the right hand movements. While watching this video I came up with three prompts to try. Been building PCs to sell and setting up my first SBC, which all feels like work so dipping my toe into the generative AI water should make for a fun afternoon. As a musician and composer, I can definitely understand the anxiety because AI can generate music too.

[@zil6470]

Thank You for this video! Great Content! It would be great if we have more videos on this topic! Thank You again and Keep Up The Good Work!

[@JJ_cl83]

@ExplainingComputers - You've made a good point about surrending control to the LLM AI algorithm. It's amazingly cool and fun to play around with but the copyright issue is a huge one, and then there are ethics and safeguards which stable diffusion has defintely not addressed at all. Until governments catch up on 'safeguards' and 'ethics', it's the wild west, especially with the uncensored stable diffusion models, especially 🤯🛡

[@bahshas]

the copyright is not that big of a deal because there already are open source ai image generators you can train yourself, its impossible to audit these with copyright laws. google has terms that they can do whatever they want with images people post using their platforms, same with facebook. funnily enough they wont get sued, they will do that to the open source ones that people can run on their own computers.

[@joefowler8464]

For someone who is artisticly challenged, this makes sense. But, I am interested in what happens in court.

[@giannistsolebas6962]

It gets more and more exciting every Sunday. Kudos! These image generators have huge potential!

[@jupreindeer9500]

History will look back upon these days that we now live in and have tons of things to say. _May they be heavier on the side of good things._

[@robinhoward9113]

impressive and at no cost.... makes you think if what you see in news or other platforms is actually real

[@sophiathekitty]

I kinda suspect that the copyright issues are going to get resolved in the most oppressive ways possible. Like major established artist and Disney own all the possible art styles and human artists can't make art without paying licensing fees for the art style they use. or otherwise resolve it in a way that works out for the existing business interests at the expense of upcoming artists.

[@SchoolforHackers]

Damn. I hope you’re not right!

[@Mightydoggo]

That´s pretty much what happened in the music industry.

[@ivanhorvat9248]

Thank you for your video, it was a joy to watch and learn. I think that we, as a society, will find ways to handle new AI technologies, as we always did with other new technologies that had come before. For people like me (without any skills in drawing, painting, composing etc.) whole new dimension is opened. I can, for the first time, create something using prompts, similar to my experience as a high-school gamer when I would create my character in certain video games. Thus, I look on this new technology as on sth good and sth that will allow me more freedom in expressing myself in field in which I could not had done it before due to my lack of skills.

[@Praxibetel-Ix]

Absolutely outstanding video! I think that the Bing Image Creator produced the best results on all but the rabbit prompt, which Leonardo AI did best. Also, THANK YOU SO MUCH for talking about how AI image generation could hurt the creative economy! I have a lot of friends who are artists that do commission work and they have had concerns about AI art. Something to keep in mind is the potential for bad actors to use AI generation to generate and spread imagery of someone as means of harassing or bullying them. Recently, news came out about AI-generated deepfake pornography of Taylor Swift being circulated on social media. I have a feeling we'll soon hear stories about young people falling victim to this kind of stuff. AI image generation may be far out, but it does have the potential of causing so much harm. Thank you very much for a great video as always!

[@billsmith1246]

I appreciated your video on AI Image Creation. I can see it might have some intriguing uses, and I might consider making a few images for my website(s). I'm not overly concerned with the implications, at least as far as the systems you demonstrated are concerned. I am not imaginative enough to get into trouble with it, and as I'm approaching 80 years old, I doubt I shall become imaginative, ever.

[@OleMose]

Interesting video, thank you Chris. I don't know how to make use of it. I'm not in that business, so maybe i can make some god images for a song at a family gathering or so. On the other hand AI is not just about images and artist. We are all going to see more of it in the future whether you are a lawyer, a graphical designer, a programmer, a nurse, doctor and so on. I totally understand the concerns, and i don't know where it all will end. On the other hand, i see it just as an other development, that makes us go further. When i was a child/young man there were a lot of business around printing newspapers. thousands of people were employed with doing typographical stuff using lead as a tool. That all changed all of a sudden and no one uses that craft any more - and yes it was also a piece of art. So things change a humans adapt and use all the new skills instead. One can only wonder if Beethoven or Handel had lived today. Would they even be able to be noticed? Of course it can be scary, but there also is a lot of new and god things waiting out there.

[@peterthepanda]

Happened just in time when Samsung released their phones that have Galaxy AI (which is powered by Google's own AI platform).

[@tedcalouri2694]

Pixel 8 also features AI camera features which brings up the question, when is a photo legally admissible as evidence? The ubiquity of phones with cameras has had a genuine impact on the justice system over the last decade. While Photoshop and GIMP have existed for years, the ability to provide images to investigators at the scene of an incident has commonly been accepted as evidence. With the Pixel 8 and other phones/cameras that manipulate images that never really existed and presents them as fact is going to distort the truth in a time where truth is increasingly difficult to come by. AI is an interesting tool, it is unfortunate that this is how it is most commonly used rather than meaningful purposes like curing disease etc..

[@s0men00bb]

In order to cure diseases (which AI already does to limited capacity) , it must be fed from patients medical records , then it crunches data like every other AI and produces (for now) very good output , in fact , sometimes AI can produce output where human doctor will have doubts , AI does not , thus proving itself for rare diseases or difficult to diagnose (due to overlapping symptoms) ones. As for the images , there's EXIF data and other forms of metadata which in turn , prove image authenticity. As for analog images , they can be tampered with too , but it gets very expensive down the line and time consuming. :)

[@adamo1139]

@@s0men00bb>As for the images , there's EXIF data and other forms of metadata which in turn , prove image authenticity. That's not true, at least not for EXIF. You can just add EXIF to an image, it's not linked to the image in any way that would remove EXIF after image is manipulated.

[@s0men00bb]

It is for modern cameras and mobiles , they apply EXIF automatically. As well as software used to edit or create images. As for tampering , nothing is totally safe , but there are dates , taken , modified , camera and software used , but digital forensics checks all that. My phone has a bunch of EXIF data embedded to the image , ofc , you can manipulate file as such , as nothing is written in the stone , but you can use EXIF for embedding and spreading malware , check for PHP/EXIF backdoor , it was big in 2013. :)

[@17th_Colossus]

As an artist I love AI, but rather than use it directly on my work I use it as inspiration. The only thing that worries me about AI is how only Stable Diffusion is open source and the rest are locked down. This is particularly a big problem because AI is constantly getting censored.

[@wartechmachine]

Great information, I didn't know there were free sites.

[@joefeely5291]

Maybe we need to choose - Let those who can afford to scoop up all the rights to "intellectual property", or make them "common" and find a "decent" way to support the creators. Tricky.

[@TheEulerID]

One thing to note is that AI generated images can't be copyrighted in the USA at the moment.

[@tonykeltsflorida]

I have made some art with AI generated backgrounds and a foreground layer using a photo from my camera. I think it works out good for personal use images.

[@dang48]

Wow! That is quite the concept. The lawsuits that have been launched are only the beginning. I think this is going to be one heck of a storm.

[@AndersHass]

It is also possible creatives use AI for specific tasks to speed up the tedious parts of the creative work. But wholesale image generation will probably be more cost effective than finding an artist for many things, thereby making it harder to earn money as an artist.

[@imjustalilbit]

As an artist, I love it. It is a new tool that gives me even more artistic freedom to create artwork in a short amount of time. This gives me the ability to structure my life in a way that allows more time for other creative endeavors that I would otherwise not have the time to accomplish. It is a great time to be alive, and I love having the time to learn about computers from this channel. Great explanation of these different tools. I am so so happy!

[@rickster2317]

Very thought provoking video Chris. While you were demonstrating the various web sites, I was thinking about all the images I could create on them. But then I began thinking about the broader implications of all this, as you so clearly discussed. As with all technology, some new tool is created, and eventually gets perverted, usually for financial gain, or because someone can. Case in point: the Internet! It will be interesting to see how Society comes to terms with this new technology. Skynet? Let's hope not.

[@jpmyers6950]

Very interesting. Original art may be a very special thing after AI takes over but like you said the day to day artists may be a thing of the past. Great video Chris!

[@wamy7619]

This was the perfect AI explanation, I wanted from Chris, thanks for the details and examples. Much of the art I that I have generated through AI, I would probably not have commissioned, my biggest issue as you mentioned with AI is where it is scraping from I would commission art specifically to what I want but that can get up there in cost. This is always going to be a challenge in the upcoming years.

[@darylmorton4611]

Hi Chris, One thing I've found is that an A I imager programme can not create all types of images yet. One example I've found, they're not good at present, is a picture of a crossbow, I've found they struggle with getting all the bits of a crossbow accurate. Fingers use to be another but they are getting better at them now.

[@christopherleadholm6677]

Hello, fellow Christopher! ...back again... Thank you for this. I have had 36 years, a literal LIFETIME of practicing "prompt engineering" using Google, as my generation specifically grew up in parallel with the birth of the internet in the early 90's. I can specifically remember using the "beige whale" my father had when I was a kid. 😂 It's amazing to see how far computers have come for the average consumer. 🤔

[@ExplainingComputers]

Greetings!

[@gst69man]

good vid. AI imaging reminds me of music creation. Every artist uses 13 notes to create music. Music is then distributed ( played through a sound device) and in various ways rewards the artist in some form, money, fame, satisfaction etc. Those 13 notes are open to anyone who wants to use them, just like public images.

[@schitlipz]

Thirteen? I don't understand.

[@jonr6680]

Except some real person created the original images they scraped. It took them a lifetime of learning to get to that point. Whereas musical notes are like building blocks that are useless without a similar level of learning skill and talent to make music with.

[@terrydaktyllus1320]

@@jonr6680 "Except some real person created the original images they scraped. It took them a lifetime of learning to get to that point." I'm sure those same artists don't give a toss if and when I go in to see my boss to ask for a deserved pay rise that I don't end up getting - my core point being that your artwork is your problem to deal with, as my salary increase is my problem to deal with and no interest to an artist. I have no sympathy for you just because you believe you're an artist - if you make a good music CD then I will buy it because it entertains me, not because you're entitled to make money as an artist.

[@gst69man]

scale has 13 notes sharps and flats@@schitlipz

[@chuzzbot]

Great video, thanks :)

[@danielrichardson5160]

Thanks Chris for giving me a whole new set of things to worry about. I've been seeing adverts about new phones with AI built in, I don't want one. As with the internet itself, it seems like a new technology that would be open so much corruption that it would soon become something uncontrollable. i liked a quote from Star Trek Insurrection " If you build a machine to do the work of a man, you take away from the man." That said, wonderful video as always. Thanks, keep up the good work and keep a stiff upper lip. Dan

[@kajurn791]

These are neat, i don't think any more professional user will prefer these web SD implementations over running it locally in their own machine for a number of reasons, but more casual/less tech savvy users can have a lot of fun with these. If you are reading this Christopher, i'd like to request a video on current AI voice generators/replacers, which i consider to have potential to be a gamechanger (pun intended) for low budget indie video game production among other things.

[@j.murphy7677]

Thanks for covering this topic. Personally, I was hoping for a prompt of Mr. Scissors or perhaps Stanley the Knife :)

[@TreeOnAHill]

Stanely was in the hand of the spotted rabbit :)

[@tzgaming207]

Excellent breakdown; thank you, Chris. I'm toying around with a video idea, just not sure which products will do what I'm looking for. Copyright issues aside for the moment, I'm looking to use an AI image generator to help design a group of characters, get them looking the way I want, & then be able to use those images to generate further images in various settings, the idea being to maintain a consistency in character design from one setting to the next. I'm guessing a paid account is likely. Anyone have any pointers? 🤔

[@mrdali67]

Can you add a normal picture to any of these AI bots as a reference say myself and get it to create eg .. a Cyborg version ?

[@ExplainingComputers]

Yes, you can provide a starting image. Of the packages shown, Leonardo AI has some great options for starting from, or adding to, existing pictures.

[@kempshott]

I hope not. It would be very open to abuse.

[@manaphylv100]

I find the img2img option in Stable Diffusion much, much more useful than txt2img, since I can just draw a rough draft in Paint (or take a quick photo) for it to base on. Its output images can then be fed back into the iterative process for as much fine-tuning as you need. Plus, I have much more control over the output when using img2img, whereas txt2img is mostly random chance. The problem with straight-up txt2img is that it doesn't understand some prompts, and you are often required to use very specific terms or spellings/hyphenations for the desired effect - better have a thesaurus ready, especially if English is not your native language. An interesting example: "twintails", "twintails hair", "twin-tails hair", and "twin pigtails" can yield very different results.

[@HKey_Root]

Superb. Thank you. Try a shiny blue spider crawling over a microprocessor.

[@smada36]

I hope that AI image generation will be to artists as the printer was to the paint brush. There will always be a want for a traditionally painted canvas, but AI will give further scope to the possibilities. It will bring mental images to the screen for those that can’t draw.

[@arvaneret_329]

Digital and quartz watches didn't vanish artisan mechanical watches. The Internet didn't vanish TV and the latter didn't vanish radio. Neither photography nor printers vanished painted artwork. I don't think generative adversarial networks will vanish the need for real, human artwork.

[@ablacknambercat]

AI doesn't create art, it creates content.

[@ExplainingComputers]

A great way to put it. Sadly true of many humans too.

[@zetectic7968]

First time seeing what AI generated images are. Very interesting video. Early days as to whether this will be a good, bad or indifferent tool and will "big tech" crush the kittle players in the market yet again.

[@lucyfrye6723]

On a related note : most people watching this channel will be able to get into AI themselves. Don't be put off by thinking it is in another realm you will never have access to because it is just too hard. The skill floor is possibly a lot lower than you might think. Play around with Python, see if you like it! Who knows if in a few months you will be training your own model?

[@ExplainingComputers]

A very good and important point. Access is widening and widely available now.

[@kerryhaycock9446]

Chris I think when you move away from your usual hardware topics , we get into some interesting areas . You should do it more often as I think it can work well , as your factual style summarises subjects so well …

[@Aruneh]

I have never used one of the online tools, but I have played at bit with software that can be run on your own PC. It's fun for a few minutes, but I don't really get the hype I must admit.

[@spiders-tours]

Nice interesting video Chris. I'd argue some of the engines are loosely AI or not AI at all. A few goes I had with SD, produced, err, junk. Covering a DB full of images with keywords and then coupling that with some fancy image merging software, is clever but not AI. The paid for and more in-depth ones, where they were actually creating from nothing a Cyborg Panda, is impressive and definitely more "AI". As others have mentioned, where I want a quick and cheap image to represent something, then these tools now meet that need without using a human. Where humans will come in is where we need more comprehensive creation or to modify what has already been created.

[@lesliedeana5142]

Although I'm definitively one that believes that one should be compensated for one's work, I have to make the argument that if I walk into an art museum and see the Mona Lisa, and I'm inspired to paint my own rendition... a week later (I'm a fast painter) somebody sees it, and gives me some Bitcoin, should I compensate someone else for my own work?

[@ExplainingComputers]

Nicely put Leslie. :)

[@simpletongeek]

You should at least cite your inspiration. So here's a question: Several people had the same inspiration as you and painted practically the same image. Should you be suing plagiarism at each other? This can happen upon popular subjects with the same prompt.

[@AppleSpaceThe]

Sure, but you are human being who was inspired. Not machine, which does not need to sleep, eat etc. It is fair competition :)

[@alanthornton3530]

That's a very good point :)

[@andrewmurray1550]

could be considered "cover art". When you consider, that musicians cover each others songs all the time. (but then maybe they have some pre-agreement with the original artist, I don't know or the OA gets a share of royalties from any income generated by the cover version.

[@Sunlight91]

In my opinion most important is the ability of individuals to produce competitive results with consumer hardware. Otherwise a few companies will dominate this new market and censor it to their liking. SDXL is very good for image creation at home, but large language models are still to resource intensive to run on high-end consumer hardware like a 4090.

[@Darkk6969]

True. It would be fascinating to see cluster servers at home for this.

[@martinsmith5028]

Thanks Chris. No doubt we could be waiting a number of years for the copyright/intellectual property rights issues to be sorted. Is there any limiting factors for the prompts? e.g. total letters or words

[@PS_Tube]

Greetings. These softwares seem useful for creating boilerplates or creating quick images for non-artists.

[@burkec33]

Yes, it will hurt artists much like cell phones have impacted photographers and videographers. We'll see a flood of "look what I produced by typing a few words" images in the coming year as adoption increases. I worry more about many not being able to discern reality from generative ai images, but this type of manipulated reality has been present in the film industry since it's inception. My biggest concern is that people will not be able to trust what they see anymore.

[@mwolfer1]

Disclaimer: This comment was created by the new, experimental "KZbin AI Responder" . The legal implication around the material used to train AI models seems to me more a distraction than an actual issue. Yes, a lot of lawyers will receive a lot of money (earn has the wrong connotation here) arguing intricate legal problems, but the drawn out process will mainly serve as a smoke screen for the actual problem you mentioned at the end. AI will fundamentally change workplaces, education, and any form of commercial enterprise, not just the creative space, similar to the way robotics changed the manufacturing space in the last 50 years, but affecting a far larger section of the working population. As an aside: This is not a surprise, Joseph Weizenbaum mentioned these implications in the late 60s and the 70s. We have been warned.

[@omnivorace]

Laurent Alexandre in France, answered to your last question my friend. And it is sad.

[@nostrumrulzz]

Thanks!

[@ExplainingComputers]

Thanks for your support, much appreciated. :)

[@jaffarbh]

Regardless of how court cases will conclude, the count down for the creative industry has already started and is unlikely to stop, I think

[@Sir_Uncle_Ned]

I can generate AI images for free using a large array of specialized models. I just run the image generation on my computer rather than on a remote server. It’s not as fast as the online options, but it works.