TTS: Molly may speak using Speech Synthesis

[Fermain]

We have TTS code and we have AI chat code, make the AI chatbot speak using TTS module.

[StianSto]

spent a little time with testing and researching this.

it takes anywhere between 9 (simplest query) and 18 seconds from sending a message to molly, to a fully writen response. most of it is writing in from stream, around 85%. the first response comes after about 1s. I have looked into the possibility of sending text as a stream, and also getting the output as a stream

• Eleven Labs API ref for TTS streaming.
• The Python way
• Example

Obviously, there will be a delay between molly writing and talking. also it seems like ElevenLabs prefer chunks, or sentences to create audio, so the audio will probably be a whole sentence or two behind at least.

if the idea is not to have a live talking bot, but like an "play message" button, then this becomes a whole lot simpler of course :)

[Fermain]

Thanks for your investigation at this time, the effort required is not worth the payoff that we will get from this feature. We will revisit in a short time once the rapidly changing world of AI has rapidly changed some more.