Imagining your characters

[MrPixel]

My mental images of characters can be summed up in these questions:

• Would I recognize the character on the street? Yes.
• Would I recognize their voice over the phone? Yes.
• Could I describe them in excruciating detail? Yes.
• Do I describe them in excruciating detail? No.
• How about summary appearance? Yes. Basic physical attributes, such as hair color and length, height/presence (tall, petite, slender, muscular).

Sometimes I'll do an image search for doppelgangers, but those are usually fruitless. AI imaging tools? So far, none have come close.

I have on two occasions chanced across someone in public who could easily be one of my characters in appearance and bearing. One time, grocery shopping, I heard a voice from the next aisle over that perfectly matched what I imagined one of my favorite characters would sound like. It was startling.

 

[ChuckWolf]

For guys in my stories, I put very little thought or words towards descriptions, other than how female characters talk about them; they're all vaguely good-looking.

Female characters are based on hyper-sexualized versions of celebrities and women I know.

My first story, Nudist Retreat Humiliation, all had women based on my coworkers and my latest story, The Making Of Widow's Journey was based on mid-70's era, Barbara Streisand, but with a strong domineering side.

 

[Kelliezgirl]

I wonder if the difference in both opinions and practice here when it comes to character descriptions is related to how people visualize things in general? There's a common visualization exercise that asks you to close your eyes and picture an apple in your mind.View attachment 2552428

If you're on the right-side of this scale you either visualize a very simplified version of an apple, or you don't visualize anything at all, and can think of an apple in purely abstract conceptual terms.

If you're on the left-side you can visualize not just "apple" but AN APPLE, and can mentally picture it in every detail, as if it were an actual apple being held in your hand.

So like, I'm definitely a one or a two. I can see the color, the imperfections, the speckles on it, feel the texture of its waxy skin, the sensation of pressing my finger into it and feeling the flesh under the skin dimple, smell, taste, everything. I CAN'T think of something without visualizing it.

I'm willing to bet that @pink_silk_glove and @yowser are ones or twos as well

And I'm not suggesting that one part of the spectrum is better than another, just different!

I'm definitely a 2, but I still think you can over describe things.

 

[Kelliezgirl]

My mental images of characters can be summed up in these questions:

• Would I recognize the character on the street? Yes.
• Would I recognize their voice over the phone? Yes.
• Could I describe them in excruciating detail? Yes.
• Do I describe them in excruciating detail? No.
• How about summary appearance? Yes. Basic physical attributes, such as hair color and length, height/presence (tall, petite, slender, muscular).

Sometimes I'll do an image search for doppelgangers, but those are usually fruitless. AI imaging tools? So far, none have come close.

I have on two occasions chanced across someone in public who could easily be one of my characters in appearance and bearing. One time, grocery shopping, I heard a voice from the next aisle over that perfectly matched what I imagined one of my favorite characters would sound like. It was startling.

I've had a similar experience. Saw a picture of someone and thought, "Yes! That's Laura!!!!"

 

[Five_Inch_Heels]

When I invoke an image in my mind of certain characters, this is an example of what I see. Impossibly perfect with 0 chance of ever meeting in person.

https://images.saymedia-content.com...NjI4MzI5/tips-on-how-to-wear-a-mini-skirt.jpg

https://i.pinimg.com/originals/b6/61/25/b661252b502eed3b3cef98fa5a7f7570.jpg

 

[pink_silk_glove]

AI imaging tools? So far, none have come close.

AI imaging is really funky. I wanted to create an avatar for my poker account. I tried like three or four different sites. I asked for brunette swept pixie haircut. That usually came out quite well. I asked 2-inch silver hoop earrings. That rarely came out well. I asked for subtle aqua eyeshadow. That always came out like ridiculous hooker makeup. I asked for a cold shoulder top. That usually came out like some totally skanky napkin blouse with the nipples poking through - gross!

Finally one site had some nice images. It would give you six options from your prompt and then you could choose one and try to tweak it from there. The tweaks didn't work very well but the original six options looked pretty good. The one that I chose wasn't terribly accurate to my prompt but I really liked it. The haircut was perfect but for some reason it deciphered brunette as peroxide blonde (??) but I do like the look of peroxide blonde and have considered doing my own hair that way in the past. The hoop earrings came out as some funky chandelier things (which I didn't mind), the cold-shoulder top came out as a black spaghetti, and the makeup came out as heavy glamor paint, but at least looked tasteful. I had also asked for a narrower angular face and that came out quite well. The bonus was that the expression on her face just happened to really work well.

AI art is fun to play with but you will never get what you ask for.

 

[Kelliezgirl]

When I invoke an image in my mind of certain characters, this is an example of what I see. Impossibly perfect with 0 chance of ever meeting in person.

https://images.saymedia-content.com...NjI4MzI5/tips-on-how-to-wear-a-mini-skirt.jpg

https://i.pinimg.com/originals/b6/61/25/b661252b502eed3b3cef98fa5a7f7570.jpg

And I suspect that if you asked multiple authors here to vividly describe these women you'd get very different descriptions and it would be anyone's guess how closely someone's mental image would match the picture after reading it.

 

[Five_Inch_Heels]

And I suspect that if you asked multiple authors here to vividly describe these women you'd get very different descriptions and it would be anyone's guess how closely someone's mental image would match the picture after reading it.

Which is why I don't go into detail.

 

[iwannadobadthingswithyou]

you will never get what you ask for

that's true, but here is how I get close. I google an image that is related to an aspect I'm looking for. Let's say "subtle aqua eyeshadow". I upload the picture in chatgpt and ask it to describe in detail the eyeshadow. then I take that description and add it to my prompt. I do this for the 4-5 main physical attributes I'm looking for and then I add the rest of the description with less detail. I end up 80% where I want. Then I can ask it to fine tune small aspects to get to 90%

 

[Five_Inch_Heels]

I also found this one recently. If she's real .....

https://i.pinimg.com/originals/a6/8a/93/a68a933520966e311a3bae2007d5aa1e.jpg


It's almost too perfect not to be machine generated. Just the depth of the colors for one.

 

[iwannadobadthingswithyou]

I also found this one recently. If she's real .....

https://i.pinimg.com/originals/a6/8a/93/a68a933520966e311a3bae2007d5aa1e.jpg


It's almost too perfect not to be machine generated. Just the depth of the colors for one.

Definitely fake, but I like your style

 

[Nynah]

Now, when I say this, please understand that I am not "creating AI art." I'm just creating some props to help me write the story. And, yeah, there's no creativity involved, or even real understanding.

I think you fundamentally misunderstand how AI art "creation" works.

Sometimes I have to ask it several times to add a detail (I gave up on placing a mole on one of my characters after the fifth time I asked for it.)

Every time you do this, whatever genAI art engine you are using scrapes pieces of art actual humans have made to create a composite image. This article does a good job of summarizing exactly what is happening whenever you prompt an AI:

It’s important to understand that OpenAI (creator of Dalle-E), StabilityAI (creator of Stable Diffusion) and Midjourney - the big three image generators - all trained their successful models by scraping millions of other peoples’ images from the internet - apparently entirely without the owners’ knowledge or permission. Lawsuits surrounding this are still ongoing.

This is worth reiterating: The billion-dollar-making generators we see today appear trained on the copyrighted works of far poorer artists, illustrators and photographers; taken directly from their portfolios and community sites like DeviantArt. This is copyright infringement on a completely unprecedented scale, and in my opinion corrupt, cynical and immoral. Paying users can even directly prompt the image generators to produce artwork in the style of an artist by typing their name - making no secret of the fact that their work was absorbed by the model.

But the pictures are there just to keep my imagination running. It's no different than the battery that keeps my laptop running or the coffee that keeps me running.

It's very different from those things. AI art is theft. Authors can find other things to inspire them. You don't need to use a plagiarism machine.

If you consider what you do to be a form of art, then it makes no sense to utilize something that steals from other artists.

There is also the environmental impact of AI image generation if you care about that sort of thing. It's incredibly wasteful on energy consumption. Laws and regulations are desperately needed around AI in general but as of right now, the effect on the environment is significant and only shows signs of increasing year over year.

Sources:
https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117
https://www.theverge.com/24066646/ai-electricity-energy-watts-generative-consumption
https://knowledge.wharton.upenn.edu/article/the-hidden-cost-of-ai-energy-consumption/

 

[iwannadobadthingswithyou]

I think you fundamentally misunderstand how AI art "creation" works.


Every time you do this, whatever genAI art engine you are using scrapes pieces of art actual humans have made to create a composite image. This article does a good job of summarizing exactly what is happening whenever you prompt an AI:




It's very different from those things. AI art is theft. Authors can find other things to inspire them. You don't need to use a plagiarism machine.

If you consider what you do to be a form of art, then it makes no sense to utilize something that steals from other artists.

There is also the environmental impact of AI image generation if you care about that sort of thing. It's incredibly wasteful on energy consumption. Laws and regulations are desperately needed around AI in general but as of right now, the effect on the environment is significant and only shows signs of increasing year over year.

Sources:
https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117
https://www.theverge.com/24066646/ai-electricity-energy-watts-generative-consumption
https://knowledge.wharton.upenn.edu/article/the-hidden-cost-of-ai-energy-consumption/

I agree with a lot of the things you're saying and I also disagree with some of the things you're saying. I'm not going to get into more detail, I don't think a Literotica forum is the place to discuss this.

 

[Nynah]

I don't think a Literotica forum is the place to discuss this.

If there's a better venue, I'm all ears. You started the topic and brought up AI art; my response was directed at that.

 

[Five_Inch_Heels]

I don't think a Literotica forum is the place to discuss this.

The fact the site policy prohibits that content due to the reasons stated says it is ... sort of.

 

[filthytrancendence]

Hey, another generative AI debate! Seems like a good opportunity to articulate some things I've alluded to in other threads but never elaborated on.

I've moderated my view on generative AI significantly in the last 6 months. I used to be a pretty extreme anti-AI activist-type. But I was convinced that this was a reductive and often counter-productive position to hold for artists. I'm not an AI-booster type, but I no longer view the use of AI as a categorical moral failure. It's a tool, and it has some problematic implications, but those implications are due far more to the economic context they exist within than the tools themselves.

If you want to dig into this deeper and listen to someone smarter and more articulate than me about it, please watch Alexander Avila's video on the topic. That wasn't the only thing that changed my mind, but it's the best version of the argument I've heard anyone make. And there's a reason it's 3 hours long. It's a very complex topic, and he makes a very nuanced case that I cannot replicate in 500 words here.

I used to think this as well:

AI art is theft. Authors can find other things to inspire them. You don't need to use a plagiarism machine.

So bear with me a second. I get the fear here. And there's been a disturbing lack of attention paid to the agency of the individual artists whose work has been used to train AI models. No argument there. That's fucked.

But the reason I moderated this view was that I could not provide a satisfactory answer to "what is the difference between training an AI and a human being learning from the work of others?"

It's not exactly the same thing, certainly. But there is an undeniable similarity. We're all heavily influenced by the artists the came before us. And that is a good thing. I think art itself is a sort of accumulated cultural resource that does cease to 'belong' to its' creator as soon as it is put out into the world. The creator certainly deserves credit for their work, and certainly they have a unique perspective on the work that nobody else could possibly replicate. They also deserve to have the value of that work recognized. But at the same time, everyone who then interacts with the work establishes their own unique relationship to the work that is no less valid. Death of the author is the concept that best encapsulates this phenomenon.

It's not that the creator has zero ownership over the work. It's that once the work is public, the effect that their work has on the people who interact with it becomes personal to each one of them in a sense that is no less real than the creator's. And this is a good thing. This is most of the value of art.

And I moderated my position on AI because it seemed to me that the anti-AI position has ended up being dragged into a much more capital-centric position on art than I am comfortable with. And this position ends up serving the interests of giant corporations like Disney far more than it does individual artists.

In summary on this point, the lines between inspiration, derivation and plagiarism is often fuzzier than any of us would like to admit. NOT always. But more often than we'd like to think. Unfortunately, generative AI training does not fit neatly into that paradigm at all, and attempting to force it over the line into plagiarism has some unfortunate and potentially destructive implications to intellectual property that are just not in the interests of individual artists. The fact is, once the training is done, asking a generative AI to create something is more like the human creative process than it is like the human plagiarism process. It is not exactly like either one, but it does not serve our interests to shoehorn it onto the side of plagiarism. Unfortunately, that argument is destined to serve the interestes of billionaires, and will inevitably end up hurting us.


This is an oft-repeated claim, but the short version is it's based on one guy's math from years ago that was reductive and faulty to begin with, and the situation has substantively changed since then as well:

There is also the environmental impact of AI image generation if you care about that sort of thing. It's incredibly wasteful on energy consumption. Laws and regulations are desperately needed around AI in general but as of right now, the effect on the environment is significant and only shows signs of increasing year over year.

Alexander does a very thorough job of highlighting the problems of these claims somewhere around the 2:40:00 mark of the video I linked. If I recall correctly, the estimates that usually get repeated are off by around 2 orders of magnitude at this point. Here (the forum won't let me link to it, but it's in the description of the video) is his bibliography, the citations for this section start at #121.

The current discourse is something like a cat fight between AI megacorportions like OpenAI and IP-holding megacorporations like Disney et all. The Disney side has thus far funded the legal challenges to AI that are obstinately originating from artists. This is not because they care about artists. It's because the interpretation of Intellectual Property required to win these legal challenges against AI companies will end up massively benefiting IP-hoarding companies. Not artists.

And ultimately, this is because in order to view art as property that a legal claim can be made against, it has to be reduced to property. The reduction of art to property is vastly more destructive to artists and culture at large than any damage that may be caused by generative AI tools, which are, at this point, spilled milk, regardless of our opinions on the ethics of their development.

 

[Bramblethorn]

It's almost too perfect not to be machine generated. Just the depth of the colors for one.

Definitely AI. The tells in that one aren't very obvious, but several other images from the same site are more blatant. For instance, this detail from a necklace is the kind of thing generative AI really struggles with:

 

[Nynah]

"what is the difference between training an AI and a human being learning from the work of others?"

Simple. AI engines are not human beings.

It's that once the work is public, the effect that their work has on the people who interact with it becomes personal to each one of them in a sense that is no less real than the creator's.

I don't disagree with this but I don't see why this matters if we're talking about AI models scraping art data without compensating the artist who created it. The effect that art has on the AI is not personal because AI is not a person.

The fact is, once the training is done, asking a generative AI to create something is more like the human creative process than it is like the human plagiarism process.

The plagiarism happens at the point of "training", which is when the model is introduced to the data set, which contains stolen art created by humans.

I don't really care what the AI "creative" process is; I'm sure it's a lot of fancy math and algorithms. And to be clear, I would not have an issue with images created by an AI if the data on which it modeled that image was truly open source and/or if the artists whose work is part of that data were properly compensated. But I have yet to see evidence that this is happening because, as the YT creator you linked pointed out, AI companies do not provide any sort of useful feedback when asked about their data models (or anything to do with their internal workings, really).

Alexander does a very thorough job of highlighting the problems of these claims somewhere around the 2:40:00 mark of the video I linked. If I recall correctly, the estimates that usually get repeated are off by around 2 orders of magnitude at this point.

I did watch that bit and I will concede there is a lot more work to be done with gauging the impact of AI on energy consumption (again, a lot of this would be easier if AI companies weren't so stingy with sharing information but oh well). A lot of his argument is centered around ChatGPT text prompts and the specifics of how much energy they consume; I did not see very much at all about AI image generation and what that equivalent is. I'm hoping there is more clarity on that in the coming years because I would be very interested to see that data.

There are several comments on the video that highlight shortcomings with his arguments, I'm not going to go through them but I would encourage you to read them because they bring up a lot of relevant points as well as first-hand accounts of how AI has harmed creatives.

The reduction of art to property is vastly more destructive to artists and culture at large than any damage that may be caused by generative AI tools, which are, at this point, spilled milk, regardless of our opinions on the ethics of their development.

I understand your points about capitalism and IP-hoarding and I generally agree with you on those. There is a lot of legal work to be done to try and make things fair and equitable for artists and not just funnel money straight into the megacorporations.

But I do not agree that generative AI is less harmful than "reduction of art to property". I know several artists who have been put out of work because publishers would rather just use AI art than pay them. A lot of artists (this includes writers as well) try to make a living off of what they create and are finding it increasingly more difficult to compete against AI art because it has become so accepted and mainstream. One of the commenters on that video made the point of how it's not just about the compensation but the feeling of invasiveness and exploitation. Even when artists create art for free, AI models take that art and commoditize it and every time an AI art image is generated it makes that shitty corporation money because you're on their website, giving them the clicks and traffic (sometimes even actually money if you subscribe to the service). If that isn't a good enough argument against using AI images (at least from a "fuck mega corpo capitalism" perspective), I don't know what is.

This is the entire thesis of my point. I know people just use AI art generators because it's fun and free. I'm merely attempting to highlight the fact that while it might be fun, it's certainly comes at a cost to someone. AI companies do need to be held accountable for what they do and how they take art and feed it into their models. But that doesn't mean people should be using these AI engines in the meantime.

 

[SkyBubble]

This is a two parter and I'd love to get your thoughts on both topics.

1. How detailed is the image of your characters in your mind? Unless they're based on real people, do you build a complete image in your head to the point that you can see their face clearly, you can imagine them smiling, walking into a room? Or is it more of a generic image, for example a blonde apparition with a crooked smile?
I try to make the images as clear and detailed as possible, I find it helps me put myself in their shoes easier. Sometimes, it even surprises me when I start with an idea and drift off into something unexpected and I find facets of the character I didn't think about.

2. Related to the point above, I've been using ChatGPT to generate images of my characters. I feed it a very detailed description that is in my head and ask it to produce an image that is usually not quite what I had in my mind. Sometimes I'm pleasantly surprised and go with the image it built, other times we go through numerous iterations to get where I want.
Do you think it would be weird to add links to these images at the end or at the beginning of the story so that the reader has an image together with the text? Or is it better to leave it to the reader's imagination? Can I even add links, does the site allow it? The images are 100% non-sexual. The characters are fully dressed, they are alone in the picture and the background may be related to the story (a pub, a party etc) or is just a generic background

I tend to give enough description that the reader can get a general idea of what the character looks like.

 

[SkyBubble]

I have a general idea of the looks of my characters, but no more than that. I never give really detailed physical descriptions. But I do always have their personalities and quirks worked out, all that makes them interesting and special.

Their smiles, smirks, teasing looks, looks of longing and desire, and especially the looks of unspoken love, those I imagine in my mind's eye when I write the appropriate scene, not before.

To answer your question, I think you should always leave some room for the reader's imagination to create something that's theirs only. I think they appreciate it.

For one of my characters, her roommate and sorority sister says she has "the best ass in the whole house." I don't go into detail. You know that she's got a great ass, but no more than that.

 

[pink_silk_glove]

Simple. AI engines are not human beings.

Agree. A human (usually) has a conscience and is aware that he may be plagiarizing. An AI bot does not and is not aware. The folks that are running the AI are profiting on a machine that they know often will be plagiarizing and somehow that is fine with them. Many AI users will generate content for commercial purpose completely unaware that the AI plagiarized something because the folks creating/running the AI apparently have no responsibility to inform users that they may be generating plagiarized work.

 

[DeMont]

This is a two parter and I'd love to get your thoughts on both topics.

1. How detailed is the image of your characters in your mind? Unless they're based on real people, do you build a complete image in your head to the point that you can see their face clearly, you can imagine them smiling, walking into a room? Or is it more of a generic image, for example a blonde apparition with a crooked smile?
I try to make the images as clear and detailed as possible, I find it helps me put myself in their shoes easier. Sometimes, it even surprises me when I start with an idea and drift off into something unexpected and I find facets of the character I didn't think about.

2. Related to the point above, I've been using ChatGPT to generate images of my characters. I feed it a very detailed description that is in my head and ask it to produce an image that is usually not quite what I had in my mind. Sometimes I'm pleasantly surprised and go with the image it built, other times we go through numerous iterations to get where I want.
Do you think it would be weird to add links to these images at the end or at the beginning of the story so that the reader has an image together with the text? Or is it better to leave it to the reader's imagination? Can I even add links, does the site allow it? The images are 100% non-sexual. The characters are fully dressed, they are alone in the picture and the background may be related to the story (a pub, a party etc) or is just a generic background

My dear IWDBTWY... I have been "writing" since a major in English and English Literature wayyyyyyyyyyyyyyyyyyyyyyyyyyy back and I can honestly say that, for me, it very much depends on the story. I begin, always, with something of a setting rather than characters since the setting really does help in forming a reader's impression of characters. Leaving some things to the readership to "fill in the blanks", in my opinion, and from what I have learned, draws them in as they invest their own imagination in the story. I am a firm believer that "spoon feeding" your audience promotes a certain boredom with the reading. Having said as much, it also depends on the POV of your tale. If you are telling it from a first person perspective then certainly, adding little "quirks" such as a "crooked smile" conveys a feeling of reality about your opposite protagonist. Even conveying that about yourself adds that touch of down to earth.

As for point 2; Generating a "picture" or "sketch" for you to keep in view as you write will help immensely in keeping your characters...er... "within character". It keeps them consistent and believable so from that point of view I believe it could be a very useful tool. As to whether you would like to add your images into the story that is entirely up to you. If you feel that your audience would gain some benefit, perhaps a better understanding of your story then it may be useful. I, personally, believe that if I can't draw my audience in with my words then I haven't done a very good writing job.

Now, all I have said is "my humble opinion", they are not rules of writing, they are not carved in stone tablets. In the end do what works for you and helps you create what you want to convey to your readers.
Deepest respects,
D.

 

[TheRedChamber]

I tend to be a 'concept and plot first' writer. Therefore my characters tend to start as a bunch of required attributes and motivations ('sexually naive', 'works in a bookshop', 'interested in the occult'). Usually only a few are needed to confirm that a real person would do the things needed to make a reasonable plot.

It's only when I start writing that secondary characteristics start to develop. The plot doesn't require her to get sloshed so whether she has a small tipple during the date or stays tee-total is something I can decide on the fly.

It's the same with appearance. I've had characters who are literal (generic-brand) Playboy models and for whom 'conventially super-attractive' is a plot driver. A larger number of my characters fall in the nerd boy/girl range and what they get upto is more important than what they look like. I often don't have much conception of actual looks on top of that and I don't think it matter *too much* in a story.

At most, it might occur to me that 'hey, Kirsty Alley would be perfect to play this character', but that is just 'play' not 'is' and Kate Bush might do just as well.

I've written twelve stories in the 'Hannah has Plans' series and all I've established so far is that Hannah is not six foot, not blonde and not overly blessed with busty substances. As the narrator is her long time boyfriend telling you what she got up to this week and not someone seeing her for the first time, I can get away with that.

Sometimes Napoleon just needs to be a short, dead dude.

 

[iwatchus]

Now, all I have said is "my humble opinion", they are not rules of writing, they are not carved in stone tablets. In the end do what works for you and helps you create what you want to convey to your readers.

I just said in another post you would fit in, but you may have the only humble opinion here. Many of us (myself included) are prone to proclaim the truth from on high. I am not sure how to cope with a humble opinion.

 

[ElectricBlue]

And I suspect that if you asked multiple authors here to vividly describe these women you'd get very different descriptions and it would be anyone's guess how closely someone's mental image would match the picture after reading it.

In my first Floating World story I never describe explicitly what my character Amanda looks like, but the depiction is based on a young woman I got to know over the course of a year. There are descriptive snippets throughout the story - which is quite long, 18k words - that cumulatively describe her. A phrase here, a sentence there, but never an exposition.

As an experiment, I asked my editor to send me some pics of women who she thought looked like my character. She sent me two or three different pics she found on the internet, and it was uncanny just how close she got to the real person.

There was a clear picture of Amanda in my mind as I wrote her, and somehow it came across in my words.