I've got the best results with BreezeTTS, Higgs Audio v3, and Fish Audio S2 Pro. audio.cpp (https://github.com/0xShug0/audio.cpp) is an easy way to run a lot of different models.
Different models have different strengths. If you throw an entire ebook at a model you're going to get a different result than if you craft a perfect 10 second sentence with a model that supports voice direction and emotion tags, so you should try a bunch depending on your use case.
Different models have different strengths. If you throw an entire ebook at a model you're going to get a different result than if you craft a perfect 10 second sentence with a model that supports voice direction and emotion tags, so you should try a bunch depending on your use case.