Thursday, August 8, 2024

AI generated Images

Images, text, video, AI, machine learning.... let's chat. 

Let's start this with some very broad statements as a lot of people freak out when you talk about artificial intelligence.  And I don't just mean uninformed or 'stupid' people as some very close friends of mine, people that I honestly look up to for their wisdom and intelligence, have overreacted to AI at least in my opinion.  

So, is AI good or evil.  I don't believe AI is even on that spectrum.  It is neither good nor evil.  This isn't the robots of Isaac Asimov nor is it the Skynet of Terminator or the Machine world of The Matrix.  AI, especially at this current level, is a tool.  Asking if it's good or evil is like asking if a screwdriver or Photoshop is good or evil.  It's like asking if the sky is 3.14... the question just doesn't make sense.  

I'm not going to go into a whole dissertation on what exactly artificial intelligence is, and I'm especially not going to try and define it and compare it against computers, programs, algorithms, and machine learning.  Suffice it to say that it's a computer doing things with less and less input from humans.  The original computers could be said to have been 'thinking' as they were doing a bunch of math computations better and faster than humans could.  That didn't make them 'intelligent' and what the AI of today is doing, in my opinion, doesn't make them intelligent.  

I think there's a big difference between 'Intelligence' and 'Consciousness'.  And I think if most people that fear AI or what AI can do, they actually fear 'Consciousness'.  A computer that is aware of itself, and thinks on it's own.  A computer that can be given orders (programs) and refuse to run them.  A computer that can generate its own ideas and work toward goals set up by itself.  THAT is truly frightening.  That is where Skynet and the Machine world and to a degree even Isaac Asimov's lovely 3 rule robots, all got the idea to control/destroy humanity.  Can ChatGPT, or Copilot, or Gemini decide to take down humanity?  No.  If it doesn't have input, it has no output.  

I don't remember the source, but I read an article where there are levels of AI.  Kind of like levels of self driving.  The big step, the one that we're all afraid of, is self generating ideas.  And we're nowhere near that.  

Now, with all of that said, I'm not saying that there isn't any reason to be concerned about AI generated content.  Whether it be text, images, or video.  All of these can be used for ill and abused.  They can even be used for 'good' purposes but have bad outcomes.  For instance, let's say I'm writing this blog post and I want an image of a woman typing at a computer.  I want the woman's right half to look like an old 3D line model, and it to slowly blend into a real photograph.  I can search and search and search all I want for an image like that, but it's unlikely that I'll find one.  But I can go to an AI generating site and I might be able to get it to make me an image.  That's fine, right?  Well, don't think of little ole me writing a blog post, and make it Apple making a print ad for the New York Times.  That's a job they'd normally have a design house do. Whether it was an internal job for their design house or an external hire of a another company, it's a team of people that would have worked to make that image.  It's creative professionals that now have less work on their plate.  Even if Apple would have went to something like Getty Images and paid the royalties to use one of their stock photos, that's still work for a photographer and graphic designer that is no longer being utilized.  Is the AI image better?  Worse?  That's not relevant at this point, it's taking food off the plate of creative individuals.  

To put a more personal spin on it, if you wanted to read a story of forced feminization, where a good guy is forced by circumstances out of his control to feminize himself into being a gamer gurl streamer where his/her sexual identity is challenged by three men and a woman in his/her life, where there's a lot of innuendo and hints at sexual activity and even some nice hot sex... Well, I think you'd enjoy reading my ongoing story Gamer Gurl.  Or, you could go to one of these AI text generators and have it write the story for you.  Going to an AI generating site might be better as it will hit on YOUR fetishes where my Gamer Gurl hits on MY fetishes (or at least the fetishes I'm choosing to focus on).  This isn't about money, but if less and less people want to read what I write, then I have less and less inspiration to write at all.  Some artists can create in a vacuum, with no input or comment from their audience at all but that's a rare gift.  Most artists feed off of their audiences reactions.  And, at least where we're at now, where do you think these AI systems learned how to do what they do?  They learned by looking at (a better way to put it is having it inputted into them) content that's already out there.  They've "read" news stories and fiction.  They've looked at photographs and art.  They've watched videos and films.  They're taking the creative energy of people and using for themselves.  But if they eliminate those very same creative professionals, who will be there to create the 'new' thing.  Who will think to put sound with moving images?  Who will think to use 3D technology?  Who will think to write out a 'Choose Your Own Adventure' style story?  Because right now, computers and AI do not generate their own ideas, so if creative professionals aren't there to copy, then we stop iterating art as a style.    

So, how do I feel about AI and it's uses right now?  Well, I don't think AI should be used to eliminate creative jobs like writing, acting, modeling, designing, programing, and stuff like that.  That being said, I think it's fine to use when it's use wouldn't have eliminated such creative professionals.  Apple using it so that they don't have to have a photographer, model, and creative team?  Bad.  Me using it to have a blog header?  Fine.  There is absolutely no way I am ever going to pay for custom work, or even use of stock images for my little blog here.  Hell, I'm a creative person myself (I want to say professional, but I don't get paid for my work so I'm not really a professional, am I?), and I use copyrighted images without permission.  I don't think it should be used in education, at least by students, now.  I don't think so because it make learning too easy, any new tool makes learning easier, but I don't believe it's reliable yet.  


Okay, I think that kind of covers AI as it stands.  I know I could go deeper into any of these subjects that I've touched upon, but I think you have enough to get how I feel about it.  What made me want to write this post out is two fold.  One, Dee has been having several posts where she plays with AI.  AI generated images and most recently, AI generated video.  We've chatted a bit about it in her comments.  More recently, a user named 'agstruggle' commented on a post of mine, asking:

"Have you ever considered adding some more AI images to your stories? The result above is stunning, and I love the story Gamer Gurl. I feel like having some images of Sadie would be a really nice enhancement."

The image they're referring to is this something I made for the second part of Gamer Gurl.  I started by using an AI site to generate this image:


The character I'm writing has distinct pink hair, large breasts, full lips, and often streams while wearing a headset with cat ears.  I have that image up on my screen as I write as I believe it's a close enough representation of that character (Nick, Sadie, or Princess BabyDoll).  As I now had that as the character, it made sense to use that image in the blog posts I was making about writing Gamer Gurl Part II.  BUT, my header graphics are wide and this image is square.  So I doubled down on the whole AI aspect and used Adobe Photoshop's 'Generative Fill' to extend the image.  I had to do it twice, as my blogs use slightly different aspect ratios (and I made the smaller one first, so I couldn't just crop it down).  As it's was generating new content both times, the images are slightly different.  Here they are with and without the title text:





I was enamored by the AI system and talked a bit about it in the posts.  And from that, I could easily see why agstruggle would like me to add more AI generated images to my stories.  

But let's back up.  I think what they want is to have more images.  If I could make them myself, I don't believe agstruggle would mind.  They don't HAVE to be AI generated.  But, like I said above, I'm not in a position to find a model that looks like Nick/Sadie, hire him/her, pay for a stylist (I can't do the styling), pay for a set or find an appropriate place to serve as a background, shoot it, do all the appropriate post processing, all just to put into a story that I write at my leisure and give away without asking anybody to pay for it.  We're talking about something that would likely cost thousands if not tens of thousands of dollars.  And that's not even considering if I wanted to do the XXX scenes as you'd pay a model quite a bit more for those.  

So, AI generated images could work, right?  And again, that makes sense... who wouldn't want to see Sadie sitting at her kitchen table while she 'roid rages out at Sarah?  Who wouldn't want to see her sitting in the car, biting her lip nervously before she goes into the lab with Todd?  Who wouldn't want to see her on the dance floor in Louis' arms kissing him, a look between hate and arousal crossing her face?  Who wouldn't want to see the (as yet, unwritten) scene where Sadie is jacking off Marley? Who wouldn't want to see Sarah sitting on Sadie's face while she uses a dildo in Sadie's ass?  

I mean... hell, --> I <-- want to see those!  

But here's the problem as I currently understand AI.  I can't make more images of 'Sadie'.  This is just one iteration that I used.  Overall, I used three different AI image generating sites.  Between them I generated around 60 images.  The text I used to generate the images centered around:

"sexy gaming woman at computer with pink hair in two high ponytails, with plump lips, wearing a headset, nervous"

And here's the image I have saved back that I was previously using as an inspiration of Sadie:


I really want THAT hair.  The color and thickness of the hair is a variable I'm willing to live with, but I WANT the twin high ponytails.  I've written that hairstyle extensively into the story.  I also want those full lips.  Along with the hair, it was the first defining characteristic of the transformation from Nick to Sadie.  He/she had those two features before he/she was dressed in any feminine clothing, had any makeup on, had a feminine voice, and even before he/she had the name 'Sadie' or 'Princess BabyDoll'.  

Here's a representation of the images I got from the AI systems:








First problem:  None these represent 'Sadie' in my mind.  None have the full lips I'm looking for.  They don't have the hairstyle either, but at least it's pink hair and I could write around the actual hairstyle.  But also notice while all the girls here look similar, it's not the same 'person'.  Her lips, teeth, nose, eyes, cheekbones, skin tones, collarbones, breasts, are all different.  They're certainly close, but they are definitely different.  And none of them have the facial expression I was going for.  The image that I ended up selecting is fine, but it's not that I was happy with it.  I was just done spending time on this endeavor.  

So far as I know, I can't save that 'model' of Sadie.  I can't take her and give her different clothes or hairstyles or makeup or backgrounds.  I can't put her with other models.  I can't put her in different situations.  

I've seen some writers/artists use AI generated content for their stories and they have models that are very close to the same between images.  But first, the images are very generic.  For example, a woman in a yellow dress looking at the camera.  Then a very similar woman wearing a green bikini looking off camera with a smile.  Then another similar woman wearing a business suit looking at the camera grimacing.   None of these images were pulled from scenes in the story.  They're close.  And I'd bet that the author got the images and then wrote the clothes and expression into the story, rather than writing freely and having an image that illustrated what they wrote.  

On the other end of the spectrum, Nicegent42 has hired an illustrator to draw images from their stories.  You can see it especially in their story Mall Bratz over at CHYOA.  Nicegent42 offered to connect me with their illustrator but I turned them down, so I don't know how this particularly works.  What I imagine is either the author explains what they'd like in an image, or they give the illustrator/artist the passage that they'd like illustrated.  

Now, why didn't I take Nicegent42 up on this offer?  And why don't I have any intent in doing so in the future?  Well, first, let me say that I think the drawings made by DreamLN (I believe this is the same artist over at Deviant Art) are good.  They match the scenes they're from and they really impact the characters affect well (happy, sad, frustrated, surprised...).  The characters also stay the same so Song and Emer always look like Song and Emer.  Sure, they change, but only because they change in the story.  What I'm saying in this first part is that I respect the artist DreamLN.  That being said, their art style isn't one that I am a particular fan of.  It's just not my bag.  I can't exactly point at what art style I think would be appropriate for my stories, but I can say that I don't think that is the correct one.  

Another reason I don't want to go down that path, is giving up control.  And no, not just from me.  From me, I'd have to collaborate with the artist.  The result wouldn't be Sadie as Caitlyn Masked sees her, it would be Sadie as Caitlyn Masked and Artist XXX sees her.  There's just no way around it.  Remember, I started out AS a visual artist, so I've been on that side of the coin.  Unlike AI, an artist isn't a tool that can be picked up and used as someone sees fit.  So, I don't want to give up that level of control.  But there's another side of control that I don't want to give up, and that's the reader's control.  

The reader controls how they see Sadie.  Sure, I'm specific that she has full lips.  But what are full lips exactly?  In real life, I have fairly thin lips, so if they were "plumped" up to what a normal woman would have, I'd consider that full.  Or are they full for a woman with already normal lips?  Or are they beyond that, as in big bimboesque lips?  Now, I don't particularly like the big bimbo lip look.  And to help, here's what I'm thinking of when I say bimboesque lips:


But if the reader wants to think of Sadie as looking like that, who am I to take that away from them?  If they think Sadie's hair is a pale pink that can almost go as blonde?  Or if they think her hair is practically neon pink?  Does she have big full eyes?  Does she squint?  Does she have a birthmark on above her lips?  These are all things that their imagination puts in.  Maybe the image an illustrator makes would be close enough, but maybe it's something that they didn't think of at all.  I can say as a fan of several authors, I'm always fearful of seeing my favorite books put on the silver screen specifically because I already know what these characters look like TO ME.  One of my favorite series of books is Stephen King's Dark Tower books and I have a particular look for the chracter Eddie.  I've recently re-read the books and realized that I have the wrong hair color in mind.  I don't know how, but I was always pictured him as having bright red, almost fake looking, red hair.  It's just how I always pictured Eddie.  But according to the book he has 'dark' hair.  Whenever they cast Eddie for the series/movie coming up, I highly doubt it will be with my imagined carrot top style hair.  

That's one of the reasons I don't want to illustrate what I write.  I certainly use images when I write as inspiration.  Like using Athena Farris for the character of Joy in It's A Man's World.  It kept me going steady with that in my head.  But I didn't write out that she looks like Athena Farris.  I didn't even get specific enough that someone could have reasonably made that comparison.  I have images in mind for Gamer Gurl that I'll share once the story is done, but they won't be 'cannon' to the story.  They're just the little bits of inspiration I used.  


That's where I sit on AI generated images.  Maybe one day the technology will get to what I need.  Where you'll be able to design a model, and then have that model put into various different situations, clothes, makeup, and have different poses and expressions.  I might even be willing to pay for such a product (probably have to pay as a service since everything is going that way nowadays).  And I think porn will help lead this charge as who wouldn't want to build up their fantasy girl and then have some perverted AI system put their fantasy girl to task blowing, fucking, sucking, and screwing all the guys and gals one could imagine?  If that could happen, I'd be really tempted as it takes care of one half of my 'control' issues.  But I'd still encourage readers to enjoy the story without the images.  Hopefully their imaginations will do the work and make something better for them.  

When it comes to AI generated stories, I haven't seen a good use for that yet.  I've read some stories by 'writers' that utilized AI generated text.  I put writers in air quotes because the stuff I've read from them before they used AI is less than appealing.  It was already bad writing without much of a distinctive style.  But I find it hard to believe that AI could generate swaths of text to fit into my story and narrative and be able to follow my tone and style of writing.  Sure, I could probably use AI to generate a scene, and then re-write it in my style... but that just seems like more work.  Why wouldn't I just do the work and imagine the scene myself, then write it down.  

I don't think people should be discouraged from using this.  Certainly there are people out there that don't have that innate writing ability and they could use the AI generated text to convey the story ideas they have.  But right now, I haven't seen AI used in this way create something that was compelling to read.  And it certainly never topped established authors ability to paint a scene or tell a tale.  

And the newest is AI video generation.  Yeah, it's neat but it's in it's infancy.  It has so many limitations and it's not generally available for people to use.  I also imagine it will be more guarded as it will be easier to create deep fakes... and no, Deep Fakes are not necessarily AI.  

No comments:

Post a Comment