Np. Im happy if someone like these lesser known characters ;)

is use at least 30 images, max around 70. u get the appearance of a char with fewer images but the lora, in my opinion will be stiffer the fewer images u use. so more different pose etc the better it works. my feeling could be wrong ofc. is used around 64 for this one, just cuz all images were small and i want min 10 for very kind of "state", with and without hair and mask and hood, outfit, nude, eyes. i clean very image and redraw parts, when for example an arm is missing or something is over the body. delete very sound effect and bubble. every other face and body. if something like that doesnt repeat 3-4 times it doesnt matter, then the ai does not learn it but, better safe then having to train it again. and i strangly like editing images ;). i dont know any guides but sure u find some here on civit ai. i was doing loras for sd1.5 before and that was much more difficult. ia asked a fried for some pointers and the rest was testen out ettings and over time i ended up with settings i liked and stick with. im sure i do it a little different then others and most likely not better and i would be nice if there were a standard in the future but it works so... if u need help if u want to do them yourself, just ask. ive read with sd3 the ai is even better at learning and it gets easier. but im afraid there comes a point when loras arent needed anymore and the ai looks up images in the web and creates chars just from that, on the other hand, that would be awesome :D