Flux knows how to create such images, but this lora greatly reduces the chance that a face will be visible
Attempting to train flux lora locally
