The biggest issue I face when generating images anywhere is that AI hasn’t been properly trained to generate "cap sleeves". The clothes always end up exposing the shoulders too much, no matter the prompt.
I took that personally and decided to fix this problem with this LoRA.
This prompt should do the trick: 4rmh013, sleeveless, short sleeves
It balances both prompts and reaches the desired result. If it still doesn't work, delete the "sleeveless" (4rmh013, short sleeves)
It's not perfect, so please send me your feedback.