The Fallen Angel image was created using NightCafe AI Image Generator. IMHO NightCafe Studio is one of the best AI image generators for newbies.
Fallen Angel
The Fallen Angel is a straightforward prompt that I copied from another NightCafe user. It’s broken down into two main parts.
First part of the image prompt:
“Creepy red shiny eyes, heavy symbiotic shape goddess, heavy black cosmic transparent wings, full body view”
Asks for a Goddess (female) with transparent wings and creepy red eyes. Also asks for a full body view. The other phrases don’t really mean much: What does “heavy symbiotic shape” and “heavy black cosmic” mean?
You’ll find with a lot of AI image prompts a lot of it makes no sense. If you were an artist how would you create an image to represent a “heavy symbiotic shape”? What is an AI suppossed to make of requests like these?
The second part of the prompt:
“intricate details, HDR, beautifully shot, hyperrealistic, sharp focus, 64 megapixels, perfect composition, high contrast, cinematic, atmospheric, moody, fractals, filigree, fairycore”
Is a NightCafe preset (Striking) with the words “fractals, filigree, fairycore” added (I added them).
I added “fractals, filigree, fairycore” to add patterns (“fractals and filigree” are types of patterns, “Fairycore” means related to fairies.
You can see the influence of the patterns in the clothing and the wings. The wings are fairy like.
Now back to the main part of the image prompt. “Creepy red shiny eyes” has resulted in red eyes, but also red eye like structures on the wings and red breasts. This is an issue with Stable Diffusion XL 1.0, it should be obvious to human artists that “Creepy red shiny eyes” is literally asking for red eyes, but the AI will generate eyes elsewhere AND use colors for other parts of the image. It can be quite annoying if you want something specific.
Next I modified the text to image prompt to:
“Creepy eyes, symbiotic demon-goddess, black cosmic transparent wings, full body view”
And:
“intricate details, HDR, beautifully shot, hyperrealistic, sharp focus, 64 megapixels, perfect composition, high contrast, cinematic, atmospheric, moody, fractals, filigree, fairycore, black and red”
The result I called The Demon Goddess of the Fallen.
Demon Goddess Of The Fallen
This changed the image quite a bit, now we have eye like nipples! We also lost the full body view.
Next I modified the text to image prompt to:
“Creepy eyes, symbiotic vampire-demon-god, black cosmic transparent wings, full body view”
I kept the other setting the same.
You will note despite changing Goddess (female) to God (male) the AI created another Goddess.
Vampire Demon Goddess Hybrid
I called this the Vampire Demon Goddess Hybrid.
I ran this prompt again and got ANOTHER Goddess (female). The SDXL AI has a strong bias towards producing female images. Generally speaking, if an image could be male or female, the AI will tend to produce a woman. It will also tend to produce and attractive/flawless young woman: it’s annoying.
Anyway, got annoyed and added Male to the prompt to finally get a male image. Next text to image prompt:
“Creepy eyes, symbiotic vampire-demon-god, black cosmic transparent wings, full body view, male”
Vampire Demon God Hybrid
Once again we’ve lost the full body view, but at least this time this image doesn’t have bright red breasts/nipples! I ran a slight derivative of this prompt again and got another male (full body view).
Lessons to be taken from this. When dealing with SDXL 1.0, if you want a man/male I suggest you explicitly mention male/man otherwise you’ll probably get a woman/female.
AI Image Creation Settings
Stable Diffusion XL 1.0 (SDXL 1.0).
NightCafe Creator model Real Cartoon XL v4: An SDXL model good at realistic looking cartoons, by 7whitefire7.
AI Image Text Prompts
Primary Text Prompt
“Creepy red shiny eyes, heavy symbiotic shape goddess, heavy black cosmic transparent wings, full body view”
Weight: 1
Secondary Text Prompt
“intricate details, HDR, beautifully shot, hyperrealistic, sharp focus, 64 megapixels, perfect composition, high contrast, cinematic, atmospheric, moody, fractals, filigree, fairycore”
Weight: 0.7
Other AI Prompt Settings
Initial Resolution: Medium
Runtime: Short
Seed: 1372517979 the seed number adds randomness to the AI image creation process.
Overall Prompt Weight: 60%
Refiner Weight: 15%
Sampling method: K_DPMPP_2M
