Can people who know Ai gen pretty well help me out? I'm trying to a gen a picture of Cynthia in bikini kneeling while bound and leashed by Jessie in PixAI, but I can't get it to keep the characteristics for each person straight. I followed the advice of that one guy from the last thread and got some good results when only using a single person in the image (way betrer than I had in the past), but the second I used a second it became a lot less accurate.

It'll sometimes switch the positioning of the characters, clothing, expressions, add in a 3rd character etc. I blew through a huge stockpile of credits I had saved today and couldn't get a single perfect or ideal image from it (there are some close ones I can photoshop to get right).
Does anyone know what my problem is, like is there something wrong with my formatting or is PixAi just shit and I have to use something paid?

The best prompt I gave was as follows, and I didn't use any LoRAs or tweak with the model or settings:

2girls,

girl 2, background, jessie (pokemon), pencil skirt, crop top, earrings, hair slicked back, eyelashes, boots, big hair, skirt, gloves, team rocket, miniskirt, elbow gloves, team rocket uniform, jewelry, lipstick, black thigh boots, thick thighs, smug, standing, holding leash

girl 1, foreground, [cynthia (pokemon)], hair over one eye, thighs, long hair, blonde hair, black bikini, collar, leash, grey eyes, hair ornament, very long hair, navel, poke ball, large breasts, tied up, hands behind back, legs tied, ball gag (poke ball), angry, blush, kneeling, beach,