I won’t go too far with the details, since my ComfyUI workflow is quite complex, but I’m using my own merge of popular realistic Pony models . Detailers were used for the mouth and eyes. Upscaling was done with a combination of upscaler models and some advanced img2img sampling.

Basic info: DMD2 SDXL 4-step lora, Steps: 12, CFG:1.5, Sampler: LCM, Scheduler: AYS SDXL. The basic prompts are listed below.

POS:

Conditioning (concat) to: 1boy, after fellatio, aqua eyes, aqua hair, armlet, bare shoulders, belly chain, bikini, black hair, blue eyes, blue hair, blue nails, blush, breasts, cum, cum in mouth, cum on body, cum on breasts, cum on hair, dark skin, ear piercing, earrings, erection, eyelashes, eyeliner, eyeshadow, facial, forehead, from above, hair bun, hair ornament, head grab, hetero, hoop earrings, indoors, jewelry, long hair, looking at viewer, makeup, multicolored hair, nail polish, navel, necklace, nipples, parted lips, (penis:1.2), piercing, pov, solo focus, swimsuit, tankini, two-tone hair, pov crotch

Conditioning (concat) from: embedding:pos/zPDXL3, embedding:pos/zPDXLrl, embedding:pos/zPDXLxxx, perky breasts, pointy breasts

NEG: Conditioning (concat) to: (3finger hand, fewer digits, missing finger, severed finger:0.1), (artist name, logo, signature, text, watermark, web address:0.1), (bikini_tan, tanlines:0.6), (child, childlike, kid, loli, teen, teenager, young:0.1), (extra digits:0.1)

Conditioning (concat) from: embedding:neg/zPDXLrl-neg, embedding:neg/zPDXLxxx-neg, 1girl