llama-tts audio streaming from LLM output generation?

[Mihaiii]

I was wondering whether llama-tts doesn't support streaming (and instead requires generating the entire audio file from the text) due to technical limitations or if it simply hasn't been implemented yet.

I mean, in the same codebase we have support for text to text (large language models) and for text to speech (llama-tts) so since llama.cpp supports streaming for text generation...maybe we can pass that stream to the tts and get an audio stream?

I'm trying to understand if this would be feasible/possible.