I had never interacted with any of the generative tools (ChatGPT, etc.) before this effort. I've been hesitant about embracing something that would cast doubt upon anything I create personally. But at the same time, people are having fun with it, and I'm a writer, not a Photoshop artist. So why not give it a shot?
I fired up Gemini this morning and figured I'd start by trying to generate some images of Abbi, the main character of my Christmas story on the Stories subforum. Taking cues from Xzzzy's Daphne Blake work, I started with Abbi from the closing scene (at Matt's family party that night):
Cute, but coy. I adjusted the prompt, adding this qualifier:Create a photorealistic image of a woman in her 20s sitting by a Christmas tree in a cozy living room. She is sitting on the floor. She has long brown hair with a white bow in her hair. She wears a black turtleneck with an ivory fur shawl over her shoulders, a long silver satin skirt, and ivory cable-knit sweater tights. She is shoeless.
Don't you just want to snuggle her?She is sitting on the floor and looking at the camera.
So I went back to the scenes at the theater. I wanted to start simple:
Gemini didn't catch my slip on "an historic theater," and it can't tell left from right, but it certainly caught the vibe:Create a photorealistic image of a woman in her 20s in a balcony seat at a historic theater. She has long brown hair with a red bow in her hair and a poinsettia flower tucked above her ear. She is looking at the camera and smiling. She wears a modest red plaid dress with red tights. Her left leg is crossed over her right. She is shoeless but there are black loafers next to her seat.
I tried again to qualify "womens' loafers" in hopes it would pick a more appropriate shoe. No dice. I then moved to the tavern setting. After a basic repeat of the theater image, I tried something more complex:
Sadly, all the attempts I got to have Abbi sit on her foot came up short. This first one also put Abbi's feet out of frame. I regrouped:Create a photorealistic image of a woman in her 20s sitting in a loveseat at a tavern. She has long brown hair with a red bow in her hair and a poinsettia flower tucked above her ear. She is looking at the camera and smiling and blushing. She wears a modest red plaid dress with opaque red tights. She is sitting on one foot and her other leg is crossed over her knee. She is shoeless but there is an empty pair of black ballet flats next to her seat.
Again, a few finer points missed, but I'll take Abbi in one ballet flat all day long. I did try to regenerate the same prompt repeating the "shoeless" element to no avail (as long as I asked for her shoes by the loveseat). Finally I ditched that qualifier and simply said "She is not wearing shoes."Create a photorealistic image of a woman in her 20s sitting in a loveseat at a tavern. She has long brown hair with a red bow in her hair and a poinsettia flower tucked above her ear. She is looking at the camera and smiling and blushing. She wears a modest red plaid dress with opaque red tights. She is stretching her legs out on the loveseat. She is shoeless but there is an empty pair of black ballet flats next to her seat.
Christmas angels come in red, too.
For fun, I tried moving her into the flower shop, first wearing her trademark black moccasins:
So we got two slippers, and not exactly the moccasins I imagined, but I wonder (and may test) what happens if I specify a brand or more specific shoe. But she was never really selling flowers in her plaid dress. So I tried her gray number from earlier:Create a photorealistic image of a woman in her 20s working at a flower shop. She has long brown hair with a red bow in her hair and a poinsettia flower tucked above her ear. She is looking at the camera and smiling and blushing. She wears a modest red plaid dress with opaque red tights. She is wearing one black moccasin slipper.
I tried prompting to have her "shoeless but standing next to a pair of black wool clogs," which generated a spare pair of shoes next to her. When I edited that to "She is shoeless," suddenly she was in footless tights and bare feet. We can't have Abbi freezing her little toes in the store. I tagged on "She is in her stocking feet."Create a photorealistic image of a woman in her 20s working at a flower shop. She has long brown hair with a red bow in her hair and a poinsettia flower tucked above her ear. She is looking at the camera and smiling and blushing. She wears a gray flannel dress and a red apron with opaque black ribbed tights. She wears black wool clogs.
After a bunch of other experimentation with some other characters from stories I've written, I returned to Abbi in the closing scene, my first attempt at this. Using virtually the same prompt as I had used (this time, specifying her looking at the camera), I got a new room and a new Abbi:
Be still, my heart.
I had figured out earlier you don't necessarily need the same prompt if you want to retain the same character in the photo. So I asked for a variant.
Behold:Create an image with the same woman, but her feet are stretched out to the fire.
Everyone should have an Abbi under their tree on Christmas Eve.
Free access to Gemini and Nano Banana gives you something like 60-75 images in a 24-hour period. That's a fair bit of experimentation. I haven't tried any proper shoeplay yet, just a couple foot-rub/massage images with other characters. Only one of my prompts was rejected as something it couldn't do, though that was largely because I was trying to trick the AI engine.
One thing that became annoying is that it stumbles a bit on changes. If I went from a character wearing shoes to not wearing shoes, often she'd become barefoot (and I only had success reverting to stocking feet once). Requests for sheer nude nylons usually got me bare feet.
More to come, for sure. But hopefully this sheds some light on prompts and what this can potentially do.
