pocketsphinx VS native-messaging-espeak-ng

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.)

I'm just trying to use coqui-ai/TTS so I can stream speech synthesis output to the browser as I do with eSpeak NG https://github.com/guest271314/native-messaging-espeak-ng. I think the issue has been brought up before on GitHub. I have not read a solution. I am ready to try again if you can suggest a minimal build process.

You can use a Native Messaging host to run local code controlled from the browser. See native-messaging-deno for a general purpose and extensible solution and deno-server where Deno's serveTls is dynamically started to run a local application, stream stdout from the application to the browser, then stop the local server.

A single page Deno server can be found here https://github.com/guest271314/native-messaging-espeak-ng/blob/deno-server/local_server.js. I have used this source code // https://github.com/chcunningham/atomics-post-message/blob/main/server.js, renamed to server.msj modified to use Ecmascript Modules instead of CommonJS, e.g.,

I've already achieved the requirement multiple ways, already; from using Native Messaging https://github.com/guest271314/native-messaging-espeak-ng, to using GNU Core Utilities tail, to Deno.watchFs() https://github.com/guest271314/fs, et al., see captureSystemAudio. The one approach I have not yet achieved is compiling to Emscripten - with SSML support.

Create a self-signed certificate. If you are on Chromium or Chrome launch with --ignore-certificate-errors-spki-list=.... Read this https://github.com/GoogleChrome/samples/blob/gh-pages/webtransport/webtransport_server.py#L42-L72. This is how I use HTTPS for Deno and Node local servers and WebTransport https://github.com/guest271314/native-messaging-espeak-ng/tree/deno-server.

I use the source code for Deno's serveTls https://github.com/guest271314/native-messaging-espeak-ng/blob/deno-server/local_server.js and wrote a Web server module for QuickJS https://github.com/guest271314/webserver-c/tree/quickjs-webserver.

This https://github.com/guest271314/native-messaging-espeak-ng/tree/deno-server is what I do using Deno

I created https://github.com/guest271314/native-messaging-espeak-ng which provides a means to send text or SSML to the eSpeak NG speech synthesis engine and parse the generated WAV in the browser. That bypasses waiting around another N years for Google to prioritize Web Speech API, which I see no evidence of Google doing - except for its cloud service.

I merely read the article to see how the author was implementing "Text-to-Speech"; how they implemented "Text-to-Speech"; e.g., native-messaging-espeak-ng to overcome or avoid the multiple issues and limitations with using the specified Web Speech API in the browser.