How could we consider conditional generaiton with DiffFlow?

I appreciate the kind open-source of DiffFlow. And I would really appreciate it if the author could answer some questions about it.
Considering CONDITIONAL generation could make it more powerful, but I'm not sure - just doing what usual DDPM does to consider the given condition could work? Or, some key issues should be taken into consideration. Do you have some idea about it?