Does anyone use A.I. images to help visualize characters?

[samhasstories]

Hi everyone,

I've been reading and writing stories on here for years (though this is my first post on this forum). I recently started playing around with some A.I. image generators, just for personal use, to stir up the imagination and help to visualize some of my characters. I was skeptical, but it's actually a lot of fun. I feel like if you get your prompts right some of their personality comes through in the images. And occasionally the images even bring to mind some aspect of my characters' appearance or personality that fit with how I envision them, but maybe hadn't occurred to me before.

Just curious if this is something any of the other writers here do? I realize the rules preclude actually posting the images, for copyright reasons as I understand it, but as a personal tool/potential creativity booster, I feel like it has some merit.

And just to be clear I use this only for images that I don't do anything with (except look at). I wouldn't use text content generated by A.I. in my stories. Not sure what the fun would be in that.

 

[madelinemasoch]

I recently made AI art of my character Eleanor from Cuckold Camp. It's not perfect but it's the closest I could get.

 

[AchtungNight]

I’ve used it, yeah. Not always pleased with the results, but it sometimes helps to visualize certain characters. I’m not posting pics.

 

[sapphyrix]

I draw my characters myself, no AI. I'm not the best artist but it's hella fun.

 

[JaxRhapsody]

No, I have quite the imagination.

 

[MrPixel]

In all honesty, I have used real photos of real people searched off the web to assist with the vision of certain characters. It's surprising how close the IRL images match what's in my head. And sometimes, not so much. AI images have not helped, as I've tried.

 

[AchtungNight]

In all honesty, I have used real photos of real people searched off the web to assist with the vision of certain characters. It's surprising how close the IRL images match what's in my head. And sometimes, not so much. AI images have not helped, as I've tried.

Done this too. The majority of celebrities in my stories have costumes based on photos I’ve seen.

 

[MyBareTorso]

I recently made AI art of my character Eleanor from Cuckold Camp. It's not perfect but it's the closest I could get.

She's SO hot! I am off to read your stories again now...

 

[Britva415]

I don't need any help visualizing my characters.

 

[madelinemasoch]

She's SO hot! I am off to read your stories again now...

I know she is. Thank you.

 

[stonedinstilletos]

I have trouble sometimes trying to visualize some of my characters. Sometimes what I’d do is go through and think who they remind of from tv or films.

What AI do people use for this though? I’m very untutored in this sort of thing.

 

[BobbyLaker]

I've had a brief dabble with AI-gen images after receiving some tips/links in another AH thread.

Found it quite interesting and will re-visit when I have some time.

Only thing was that my attempts at 'sunbathing women' always seemed to end up with amputees or people with unfeasibly configured limbs. And no, I didn't include the word "tentacle" in the generative terms

 

[samhasstories]

I have trouble sometimes trying to visualize some of my characters. Sometimes what I’d do is go through and think who they remind of from tv or films.

What AI do people use for this though? I’m very untutored in this sort of thing.

I've dabbled with a few. I've enjoyed the results I've gotten from promptchan AI, which is sort of a specifically NSFW version (which suits my purposes, at least). Unfortunately after a few freebies they want you to buy credits. Another I've found that gives you a fair amount of free use is tensor .art, which is a little more confusing to use, and results I've gotten have been uneven. But might just be a learning curve - there are a lot of different 'models' to try out on there.

I've had a brief dabble with AI-gen images after receiving some tips/links in another AH thread.

Found it quite interesting and will re-visit when I have some time.

Only thing was that my attempts at 'sunbathing women' always seemed to end up with amputees or people with unfeasibly configured limbs. And no, I didn't include the word "tentacle" in the generative terms

I've gotten some unexpected body shapes, too - extra arms, backwards hands, weird floating limbs, demon eyes... Definitely some throwaways, but it's generally a decent balance with 'realistic' results.

 

[AchtungNight]

One weakness of AI I’ve found is it doesn’t have a large amount of imagination. It can’t envision actress Erika Christensen in costume as video game heroine Joanna Dark, for example. Only one character or the other. At least not any program I’ve found. Sometimes you have to list characteristics in the right order also. And remind the AI in some cases to get hair or clothing color correct. Erika would need red hair or a wig to impersonate Joanna, as she is naturally blonde. Be sure to tell the AI to give her red hair! Listing each item of the costume’s clothing individually works for some programs, but more often, inevitably, they add a weird facial expression, six fingers, or other disturbing things that scream “I don’t understand how to depict humans!”

Nonhuman characters are much simpler for AI. Sometimes. Still, for certain programs, you need to make clear what you want. “Vampire Scarlett Johansson, red dress, display fangs” for the other FMC of my Halloween story.

Style is important too.

 

[tenyari]

Seeing this topic got me to look for my past posts on the subject and I've found that all the old threads where some AI art had been posted seem to be gone from the forums here. No idea what happened - but if they've decided to ban AI art here, those posts up above might want to edit their content.

 

[MillieDynamite]

Who has time to write the script that creates these images to inspire them to write a story they could be busy writing rather than making AI and tweaking it until it inspires them?

 

[MillieDynamite]

Well there ain't nuttin wrong with lesbians.

My prompt was literally auburn haired woman with green eyes (I have a type...), and dark haired guy in business attire. I just hit generate bunch of times and now I have a cache of characters to choose from, including a lesbian pairing.

 

[samhasstories]

Yeah my prompts are pretty simple. A tweak here, a tweak there. As far as time goes it's worth it - and honestly it's more for fun than inspiration, though it does get the imagination going. Not that it needs much prompting.

 

[tenyari]

In the hopes that this thread survives.

I recommend installing Stable Diffusion or another Ai Art tool on your computer so you have the full suite of tools to improve an image.

The power of it comes into play when you start to take your work back and forth between the Ai Art tool and Gimp/Photoshop. Using 'InPaint' or 'Img2Img' to refine an image.
InPaint is a tool where you mask out an area and make new prompts to have it redraw that area according to your new prompts.
Img2Img is where you load in an image and have it modify that image based on your prompts.

These are tools in stable diffusion, and I suspect other AI Art applications have similar equivalents.

The best workflow is to make an image, take it into an art tool like Photoshop and remove things, manually draw in some new stuff, blend images, and so on - then take it to Img2Img and clean up your changes, merging them in smoothly or even upping their quality (turn a stick figure into a realistic person in that spot and pose).

My process from prompt to final image can involve 100 or more passes through the tool, with a similar set of manual drawing runs in Gimp.

AI Art, done artistically; is more "hand made" than most real world photography. But due to controversies over the original training methods it has no respect as of yet and can't even be copyrighted whereas a photo of something that already exists can.

 

[CyranoJ]

Plenty of inspiration to be taken from IRL models and non-AI art when needed, for me. I don't feel the need to engage with AI art -- much of which looks subtly "wrong" to me -- and its various controversies. That might change if the technology becomes more ethical and the intellectual property issues get resolved... but I suspect not.

 

[samhasstories]

In the hopes that this thread survives.

I recommend installing Stable Diffusion or another Ai Art tool on your computer so you have the full suite of tools to improve an image.

The power of it comes into play when you start to take your work back and forth between the Ai Art tool and Gimp/Photoshop. Using 'InPaint' or 'Img2Img' to refine an image.
InPaint is a tool where you mask out an area and make new prompts to have it redraw that area according to your new prompts.
Img2Img is where you load in an image and have it modify that image based on your prompts.

These are tools in stable diffusion, and I suspect other AI Art applications have similar equivalents.

The best workflow is to make an image, take it into an art tool like Photoshop and remove things, manually draw in some new stuff, blend images, and so on - then take it to Img2Img and clean up your changes, merging them in smoothly or even upping their quality (turn a stick figure into a realistic person in that spot and pose).

My process from prompt to final image can involve 100 or more passes through the tool, with a similar set of manual drawing runs in Gimp.

AI Art, done artistically; is more "hand made" than most real world photography. But due to controversies over the original training methods it has no respect as of yet and can't even be copyrighted whereas a photo of something that already exists can.

That's interesting, thanks tenyari, I'll make a note of your advice.

I think it's an interesting question, too, about what constitutes a legitimate piece of work, based on the relative technological sophistication of the tools used to produce it. If you're inputting that much of your own manipulations it would be hard to argue that the end result isn't "yours." But I guess that's a whole other can of worms.

I did delete my previous post that contained one of my results, just in case that is against the rules.

 

[Codoc24]

In the hopes that this thread survives.

I recommend installing Stable Diffusion or another Ai Art tool on your computer so you have the full suite of tools to improve an image.

The power of it comes into play when you start to take your work back and forth between the Ai Art tool and Gimp/Photoshop. Using 'InPaint' or 'Img2Img' to refine an image.
InPaint is a tool where you mask out an area and make new prompts to have it redraw that area according to your new prompts.
Img2Img is where you load in an image and have it modify that image based on your prompts.

These are tools in stable diffusion, and I suspect other AI Art applications have similar equivalents.

The best workflow is to make an image, take it into an art tool like Photoshop and remove things, manually draw in some new stuff, blend images, and so on - then take it to Img2Img and clean up your changes, merging them in smoothly or even upping their quality (turn a stick figure into a realistic person in that spot and pose).

My process from prompt to final image can involve 100 or more passes through the tool, with a similar set of manual drawing runs in Gimp.

AI Art, done artistically; is more "hand made" than most real world photography. But due to controversies over the original training methods it has no respect as of yet and can't even be copyrighted whereas a photo of something that already exists can.

I've recently started using Easy Diffusion on my laptop. Mostly just to play around with and see how it works.

In the hopes that this thread survives.

I recommend installing Stable Diffusion or another Ai Art tool on your computer so you have the full suite of tools to improve an image.

The power of it comes into play when you start to take your work back and forth between the Ai Art tool and Gimp/Photoshop. Using 'InPaint' or 'Img2Img' to refine an image.
InPaint is a tool where you mask out an area and make new prompts to have it redraw that area according to your new prompts.
Img2Img is where you load in an image and have it modify that image based on your prompts.

These are tools in stable diffusion, and I suspect other AI Art applications have similar equivalents.

The best workflow is to make an image, take it into an art tool like Photoshop and remove things, manually draw in some new stuff, blend images, and so on - then take it to Img2Img and clean up your changes, merging them in smoothly or even upping their quality (turn a stick figure into a realistic person in that spot and pose).

My process from prompt to final image can involve 100 or more passes through the tool, with a similar set of manual drawing runs in Gimp.

AI Art, done artistically; is more "hand made" than most real world photography. But due to controversies over the original training methods it has no respect as of yet and can't even be copyrighted whereas a photo of something that already exists can.

I've recently started using Easy Diffusion on my laptop. Mostly just to play around with and see how it works. I became annoyed with some of the AI websites that would tell me I was trying to create NSFW content when I wasn't. One of them told me my content was NSFW when I prompted for a couple on a yacht wearing swimsuits. That was the proverbial straw, so I downloaded and installed Easy Diffusion. Now I can create as many images as I want and not have to worry about paying anyone or getting blocked for NSFW.

I created one image that I used for a cover for a book on Smashwords. I make very little money on Smashwords, and if I pay someone to create a cover, I make ZERO dollars. I'm not trying to make a living off it; writing is mostly a hobby.

 

[MillieDynamite]

AI is a tricky subject for publications. If you use it to create (shame on you) text for a story, that's verboten. If you use it to spur your imagination, that's fine. At Amazon, if you use it to write the story or part of the story, you have to tell them. However, if it's used to help inspire your juices, you don't even have to tell them. The thing is, they aren't banning AI but want you to tell them if AI creates part of the story. However, if you use it as an editing tool, it's okay, and you don't have to tell them. If cover art is created by AI, you have to tell them. If cover art is enhanced with AI, you don't.

I'm glad I'm not the one actually posting them, and it's the publisher's problem. I do think that some of the covers I've had were AI-assisted.

 

[vanmyers86]

No, I generally start with vague images of the characters, then keep writing until I can see them. However, if their physical appearance informs their character and/or moves the plot in some way, then I include it up front.

I prefer to keep appearances general, so the readers can envision what they personally like. But if generating images helps you write, keep doing it!

 

[LonRivers]

There is a wealth of imagery on the internet to use. Books and magazines offer tons of photos in which to spark your imagination. I don't use AI for that. I can see an ad for shampoo and the model sparks an idea in my head. Or I can watch a television show and the character says or does something that has my creative juices flowing. There is just too much out there already to use AI. BUT, if you are trying to narrow it down to a specific subject or idea, then AI is a good place to start. Just keep in mind that AI doesn't mean you JUST came up with it. I have seen AI come up with pictures that I have seen on other websites. It's not all new material. Some of it is artwork protected by copywrite. Choose your AI wisely.