Found the cause in a staled environment of the service. Text-to-Speech generation is now functional again.
We currently see that new Text-to-Speech generation fails. Investigating