It's less about the prompt and more about the process. I generated a shitload of images at a lower resolution (it's very fast and works in batches of 8 on my 3070), and then sent the ones I like to img2img, with the same prompt and upscaled them 2x.
Prompt
(nsfw, 1girl, 1boy:1.2), ((foot worship)), ((naked woman with (green eyes), (short brown hair with green tips), smirk, ((nipple, pussy, pussy juice)), (dripping cum from the pussy), (tan)), masterpiece, best quality, brightly lit hall, detailed background, (pov), from below, Tonee, Andreas Achenbach, Boris Vallejo
Negative prompt: [3d], mosaic, black and white, censorship, trap, (large breast), (((ugly))), ((duplicate)), ((morbid)), ((mutilated)), ((transsexual)), extra fingers, ((poorly drawn hands)), ((poorly drawn face)), ((deformed)), blurry, ((bad anatomy)), (((bad proportions))), ((extra limbs)), (((more than 2 nipples))), (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), (too many fingers), worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name
Steps: 48, Sampler: DPM++ 2S a Karras, CFG scale: 7.5, Seed: 758727136, Face restoration: GFPGAN, Size: 640x448, Model hash: 38c1ebe3, Batch size: 8, Batch pos: 4, Clip skip: 2
It's less about the prompt and more about the process. I generated a shitload of images at a lower resolution (it's very fast and works in batches of 8 on my 3070), and then sent the ones I like to img2img, with the same prompt and upscaled them 2x.