I'm not entirely sure what the problem is, but make sure you're using my prompt, negative prompt, model, resolution, sampler, seed, steps, noise scheduler, cfg(guidance scale), and embeddings. Those things matter a lot in terms of reproducibility.
Also keep in mind what I post here is generated first through txt->img until I get something looking similar in form to what I'm envisioning, then edited by me using many inpainting steps sometimes combined with a little drawing by hand to get things in the ai refuses to draw (like Jessica's hat), fix ai nonsense, change character details, and so on.
So the result you see is actually a combination of the original prompt generation and a lot of effort in my part (2-5 hours). For example, if the image you get for the Jessica image prompt settings looks like this:
View attachment 3784248
Then you've used the settings correctly to get the txt to img generation I started with. If it doesn't, then you did something wrong.