I have raspberry pi. I don’t currently use any TTS, I download the direct audio on Amazon Polly (TTS) and then I convert it to audio and play it when I ask my rhasspy assistant a question.
My question is: Is what I do feasible or do you recommend using direct TTS?
I want to use one that looks like a human voice. I think Mozilla’s does it, but I don’t know if it’s already available for Raspberry Pi.
Why not use Google Wavenet or Larynx?
I want to work offline (English and Spanish). I tried one that sounds robotic, I don’t know if it was Larynx.
Because if it sounds robotic I prefer to play a previously recorded non-robotic audio.
What do you think?
Check out the Larynx voice samples to see if it’s good enough for your purposes. I’m currently working on version 0.5, which has some decent performance improvements. Larynx is still slower than other offline TTS systems, but I think the quality is much higher.
With Google Wavenet, only once the text is send to the cloud. After that it is played from cache. So basically what you do now with Polly, only then automagically
The majority of your response is probably a small set, so after a couple of commands all is played from cache.
Larynx is also good 