Kokoro gives pretty good voices and is quite light...making it useful despite its lack of voice cloning capability. However, I haven't figured out how to run it in the context of a tts server without homebrewing the server...which maybe is easy? IDK.