trained on around 17 images from midjourney of characters in studio.

This is meant to re-create the concept of recording in Studio.

In the training data, there was a lot of emphasis on smoky rooms, gold chains, gold teeth, etc. so you may want to implement those in your promise. It could be a bit heavy handed so I don't know if I would have the weighting set to high try at a lower value around 6 first and work your way up.