Now we come to my favorite part of this Summer of Narnia: AI depictions of Jadis.
First a blast from the past. The picture below is an authentic photo of a Victorian actress whose “exotic” costume is more inspired by Western takes on non-Western fashions than anything authentically Middle Eastern. On the right, Pauline Baynes’ version of Jadis from The Magician’s Nephew. See how alike they are?
Now what happens when you combine these two, in Midjourney, with some line drawings of similar ladies?
Voila! A quartet of Jadis clones who may have come from the same well of inspiration.
Taking it further, I cobbled my own “line” drawing and combined it with a pen-and-ink rendition of Roman ruins by one of the Old Masters, again using Midjourney’s blend command.
My model was a Bollywood actress in a sari and the same Roman ruins. Maybe in a later post I’ll go into how I combined the two. Not bad, right?
Sessions in NightCafe came up with the following versions of Jadis using their new SDXL 1.0 model, which I highly recommend for generating a range of human expressions.
“Aztec Egyptian Chinese Empress, long black hair, angry face, looking down from the balcony of her palace, watching the city burn below, red setting sun.”
Well, there’s no city or red sun. But I like her impassive yet sneaky expression and the palette, which adhered to adobe-hued reds, blues, and golds. On its own the program gave her a facial scar. Though her costume is not symmetrical, and her fingers are tangled, the aesthetic works. It’s a combination of many ancient civilizations yet looks like none of them. The texture is also marvelous, very matte, like dry gouache, just the tiniest bit less than perfect as if painted by hand.
“Aztec Egyptian Chinese Empress, long black hair, angry face, looking down from the balcony of her palace, watching the city burn below, red setting sun.” This time I used the pic at the left for a figure reference.
And, wow! I love Jadis’s proud, fierce expression here, and her costume, and the suggestion the city is burning. It’s not perfect — why are there two red suns? — but this was very, very good. On the downside, there’s still a hand problem: multiple fingers, banana fingers, hands too long or too short. The one above was the best of the bunch. Even so, I cropped it to get rid of some of the background wonkiness.
With some work on Photoshop, this would be very professional. “Some” meaning more than a few hours. But I’d do it for a self-published cover.
How does NightCafe’s SDXL handle a photorealistic image? Say, Jadis as she was found in The Hall of Images?
“Aztec Chinese Sumerian Empress, long black hair, about to sit down on a carved stone chair in a dark palace hall, to the left and right are many other kings and queens also seated.” I used the Balinese model to the left.
The second half of the prompt was lost, but the image is compelling, a sort of alternate-universe Olmec aesthetic save for that odd pocket in the bottom of her gown which might house some sponges. There’s also the unevenness of the throne’s arms – one side is thicker, lower, and extends forward more than the other one in an Escher-like visual dissonance. Probably I should have used a seated model not a standing one. And again, the hands are deformed. But overall I was impressed at what the program could do, especially with the variety of faces, crowns, and headdresses.
Blending a number of lovely Indian, Chinese, and Malaysian costumed ladies in Midjourney gave me this Jadis, resplendent in gold, red, and bright yellow.
“Who has rung the bell and awakened me?”
She looks very real, but there’s also less character than there was in the NightCafe images above. She’s merely a model that’s posing for a magazine, with a few inconsistencies like the missing shoulder pad on her right shoulder. Costume-wise these were all pretty delicious, but not really Jadis.
What if I ran some of the NightCafe pics through Midjourney as image prompts?
Not bad, but again, the marvelous character of the originals was lost. The queen on the bottom one has seriously wide Kim Kardashian hips.
Another Midjourney creation. This was not my prompt but another user’s who asked for a photorealistic version of Zendaya as an Egyptian/Japanese goddess in a cyberpunk Tokyo. I like the hyperdetail of it and that it looks neither Ancient Egyptian or Japanese but something new.
Here’s a surprise! Another of my Midjourney blended Jadis images used a pic of a Phillippino Miss Universe contestant, and on its own the program gave her a red panda motif for her gown. I wanted to show her stealing the Silver Apple from the Western Garden but she’s merely posing.
Now let’s see how StarryAI handles Jadis.
To the left are are the two of the previously generated pics I used for reference. I used the Photos method and Cinematic style, which gave me the above pics with the prompt “Aztec Egyptian Chinese Empress, long black hair, angry face, looking down from the balcony of her palace, watching the city burn below, red setting sun.” I am super pleased with these results — run through several generations, I got a more believable and exotic character who looks defiant but also full of regret. The seated Jadises, however, were as full of wonkiness as their originals.
So… what an interesting experiment! I learned a lot, and it was fun.