It was the first time I started using the 'oscillating inputs'. Other tries haven't come out quite as good, but it's an interesting concept. It seems usually the first one wins.
It is, yeah. I just use dynamic prompts to combine a bunch of different random values, and some things come out pretty fun.
((RAW Photograph)) ((NSFW, half undressed)) A wrestling 50 year old chubby pakistani empress wearing an abaya In a body modification chamber art, elemental, (graffitti:0.98)
lora:Elemental_Graffiti-step00001500:1 lora:d\EnvyZoomSliderXL01:-0.7 lora:add_detail:1 Negative prompt: ((malformed)), ((worst quality)) Steps: 55, Sampler: DPM++ SDE Karras, CFG scale: 9.5, Seed: 3651687541, Size: 1024x1024, Model hash: 69cc62d2b6, Model: newrealityxlAllInOne_30Experimental, Lora hashes: "Elemental_Graffiti-step00001500: 88c39f50d2ab, d\EnvyZoomSliderXL01: ed47526bdf74, add_detail: 7c6bad76eb54", CodeFormer visibility: 0.442, CodeFormer weight: 0, Version: v1.8.0
These are absolutely fantastic, thank you for sharing them!
Thanks for posting these! I wish I had a similar list to contribute, but I very much appreciate what you're trying to kickstart here.
parameters
A gorgeous full body photograph of a naked african woman getting married in a church. Extreme tattooing. Negative prompt: 4 girls high contrast, (deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime:1.4), text, close up, cropped, out of frame, worst quality, low quality, jpeg artifacts, ugly, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck Steps: 40, Sampler: DDIM, CFG scale: 9, Seed: 3830865063, Face restoration: CodeFormer, Size: 768x512, Model hash: f2f43201f3, Model: kitchensink2nsfw_, Denoising strength: 0.5, Token merging ratio: 0.5, Token merging ratio hr: 0.5, Hires upscale: 2.5, Hires upscaler: ESRGAN_4x, Version: v1.3.0
I love that that prompt produced the lower stomach tattoo! It's definitely a very sexy effect overall.
You're exactly right. The thing to do in this case would have been to img2img inpainted it until the braids matched, but I didn't this go around. Weird hair is definitely one of the tells though.
That was entirely the output of the model, the only thing I did as a human was img2img it to make one I liked larger. The split screen effect was entirely due to this lora: https://civitai.com/models/42214/onoff-lora
In simple terms, it's a small modification to a model which usually produces some kind of specific change, commonly an art style, a pose, a hairstyle, something like that. This link describes it in more depth: https://softwarekeep.com/help-center/how-to-use-stable-diffusion-lora-models
You can find lots of them here: https://civitai.com/
If you see a 'Lycoris', it's like a fancier lora.
I'm glad you like it! I think there are some nipple focused loras on civitai though mostly for very flat women (which I also like, but wouldnt work in this case). Ill see how they go.
Yep, even with the pussy focused lora. I could probably have gotten some examples had I kept trying with different language, but I liked this one well enough.
Here's with some added tattoos and a similar prompt:
This one is straight up amazing
That's a little funny, as I wouldn't say it's especially backlit. Though the model interprets things the way it does, so hey.
Tried it with german instead of libyan and wound up with one which didn't call for any inpainting:
It's a real pain to use correctly, but I think the regional prompter extension to automatic1111 is currently the only way to do that: https://github.com/hako-mikan/sd-webui-regional-prompter
Which is to say, I've found 'just give different prompts to the left and right' to be ok-ish, but I've had real trouble in setting up a complex 2d field of different prompts.
I'm I'm wrong on either front, please someone let me know!