Yep this can be normal because i use this system of process in Leonardo!
(free for 150 "tokens / Day" --- failures include so normally 100 images /day with low regulate/size)
I use a "training model" made in Leonardo with only 5 images made from "Wombo" (not well size trimmed, positions, number, styles etc... i was beginner
And alas training models building is not more allowed with Leonardo free version)
But this early one works very fine!
The Leonardo data set ( as you can see i should add 2 black sides for have a 512*512 size !
(beginner again)
And after
any image as start image...
So a sort of kingdom of randomness!
Advantage : no elaborated prompts at this date... something like "portrait" ,"furious portrait" or "smiling portrait", "landscape", "form" etc... is sufficient!
And no time losted...life is to short!
Have fun!
PS Leonardo has the more easy Data set Engine building that i have seen!
PS of course i could modify the result (even in AI ) for maybe a more cool result but it's not my present wish...i can make this in any time in the future if needing!
It's that i made since 2 series i take old images from Wombo (an another cool free AI) where eyes were bad and not elaborated...Leonardo makes them better in variations without effort!