pocketsphinx VS GoogleNetworkSpeechSynthesis

There are espeak-ng https://github.com/espeak-ng/espeak-ng and pocketsphinx https://github.com/cmusphinx/pocketsphinx which can be used locally without making external requests.

And so I installed PocketSphinx, "one of Carnegie Mellon University's open source large vocabulary, speaker-independent continuous speech recognition engines."

- https://github.com/cmusphinx/pocketsphinx

This avoids having to stream audio 24x7 to a cloud model which would be super expensive. This being said, I'm pretty sure what the Alexa does, for example, is send any positive wake word to a cloud model (that is bigger and more accurate) to verify the prediction of the local wake word detection model AFAIK.

PocketSphinx is generally regarded among voice assistant communities as a less reliable, but straight OOTB, alternative to a robust listener. It's a good solution when you want multiple hotwords (or just aren't in a position to train even one word.)

If you want you can use Google's TTS service without Web Speech API with a proxy, see GoogleNetworkSpeechSynthesis. I made a feature request for Google to release the source code as FOSS Re: Issue 263510047: Release TTS and STT source code and Google voices as FOSS so we don't have to make external requests just to use Web Speech API.

I requested to Google to Release TTS and STT source code and Google voices as FOSS which you can request over the network here GoogleNetworkSpeechSynthesis. Those are the voices Google Chrome uses for Web Speech API. Feel free to chime on the feature request in in support of Google releasing the source code of its network-based cloud service (that google uses for Web Speech API implementation) TTS and SST code as FOSS.

Technically you can use Google API key shipped in the browser see guest271314/ GoogleNetworkSpeechSynthesis . No need to sign up for a Google account to use this. I don't think Google's TTS is related to "AI".

This is how you can make the request yourself GoogleNetworkSpeechSynthesis.

If anybody want to build a TTS "module" or "component" that accepts and processes SSML input, and that the user can control - to an appreciable degree - on the front-end you can play around with this https://github.com/guest271314/GoogleNetworkSpeechSynthesis. (Makes an external request). Have fun!

It is possible to use Google Network Speech Synthesis service "free of charge", see GoogleNetworkSpeechSynthesis. On Chrome if you use Google voices with SpeechSynthesisUtterance() and window.speechSynthesis.speak() you are doing that anyway.

I implemented SSML parsing in JavaScript by hand for Web Speech API per SSML specification https://github.com/guest271314/SSMLParser, where the Web Speech API nor Firefox nor Chrome or Chromium browsers (Google does implement SSML parsing as a service https://github.com/guest271314/GoogleNetworkSpeechSynthesis) have implemented SSML parsing.

FYI when Google voices are used on Chrome or Chromium-based browsers the browser also makes a remote request, see https://github.com/guest271314/GoogleNetworkSpeechSynthesis.

How is processing SSML an attack vector? Google provides that capability as a service - with embedded limitations on character input that are not specified. This can and should be implemented in the browser - without making an external request; native-messaging-espeak-ng, GoogleNetworkSpeechSynthesis.