Rendered all examples I could Seems the model has issues with detailed voice representation, but guess that's just architecture limitations at this point