well this is a general problem of Stable diffusion.
try this here: https://stigmavlc.github.io/lora-cast/

from the creator:
Built something I want to share: LoRA Cast A problem I kept seeing people run into here — you train a face LoRA, it looks great on its own, then you try to put two people in one picture and their faces blend into each other, or everyone ends up wearing the same face. Masking and inpainting help a bit, but the LoRA is still loaded for the whole image so it never really goes away. LoRA Cast never renders two identities together. It finds the faces, you say who each one is, and then it redraws each face on its own — one render, exactly one LoRA — and blends it back with colour matching, a feathered edge and matched grain. Two LoRAs that share a trigger word can sit in the same picture without a problem. It runs entirely on your own machine on top of ComfyUI or Draw Things. No account, no uploads, nothing leaves your computer. Free and open source. There's a demo on the site where you can click around the actual interface without installing anything. https://stigmavlc.github.io/lora-cast/ This is a beta. It works, but it's barely been used outside my own setup, and Windows and Linux are completely untested. So please install it, throw real work at it, and tell me what breaks — bugs, rough edges, confusing wording, all of it. That's exactly what I'm after right now.