For my usecase, fish audio voice clone funny enough is actually superior in humanness as well as accuracy, the only problem is that on any long call, the audio generation fails and has to fallback to my backup voice. This causes the bot to repeat what it said, hence ruining the experience.
Is it at all possible to somehow solve this. It must be having trouble calling the fish audio endpoint. Really makes it hard to have faith in running 50k calls per month