Thanks! Try the --raw-stream option for listening to long texts: https://github.com/rhasspy/larynx#long-texts

For speech-dispatcher, I'd start a Larynx HTTP server and use curl to get audio. I have an undocumented --daemon flag that does something like this.