A JavaScript speech synthesis and visualization system for the web, based upon meSpeak and sam-js.
Jaxcore Say combines both speech synthesis systems through a common API with an HTML5 canvas-based audio volume display which was partially based on volume-meter.
Beware that web browsers are now disabling Web Audio API by default, so the first usage of Jaxcore Say must be called from a user action such as mouse click or keyboard action.
For a version of this library which works inside NodeJS, refer to:
CyberTyper is a ReactJS component for automatically typing tech synchronized with jaxcore-say's speech synthesis
- Example: Blade Runner interrogation
npm install jaxcore-say
English en is the default, but see below for how other languages can be used.
import Say from "jaxcore-say";
Say.setWorkers({
'espeak': 'webworkers/espeak-en-worker.js',
'sam': 'webworkers/sam-worker.js'
});
let voice = new Say({
profile: "Cylon",
language: "en"
});
voice.say("hello my name is Cylon").then(function() {
console.log("done");
});
Both speech synthesizers are loaded with webworkers which drastically improve memory usage. The webworker files can be downloaded here. Or to build the worker files clone this repo and run npm run build-workers and see the dist directory.
The worker js files must be served separately from the NPM module, and the path to the files must be set using the Say.setWorkers() method.
SAM is English only so there is only 1 webworker file for it.
But ESpeak has many language options. For English (US) only use this:
Say.setWorkers({
'espeak': 'webworkers/espeak-en-worker.js',
'sam': 'webworkers/sam-worker.js'
});
For other languages it likely will be better to use the all file:
Say.setWorkers({
'espeak': 'webworkers/espeak-all-worker.js',
'sam': 'webworkers/sam-worker.js'
});
Keep in mind, the file sizes of each worker is large. English (US) is 1.7MB, and all the languages is 2.7MB. Your webserver requires proper caching of the worker files.
Jaxcore Say includes the following predefined ESpeak-based voices:
- Jack
- Pris
- Roy
- Scotty
- Xenu
- Cylon
- Leon
- Rachel
- Zhora
- Borg (a combination of other espeak voices)
And the following SAM-based voices:
- Sam
- Elf
- Robo
- Granny
Custom voices can be generated at run time. See the source code of SayApp.js example for details.
The voice profiles include an easy way to modify the speed (faster/slower) and pitch (deeper/higher):
voice.say('hello world', {
fast: true
});
voice.say('hello world', {
slow: true
});
voice.say('hello world', {
high: true
});
voice.say('hello world', {
low: true
});
The language should be defined while instantiating new Say(), the possible languages are:
- ca = Catalan
- cs = Czech
- de = German
- el = Greek
- en = English
- en-n = English (N)
- en-rp = English (RP)
- en-sc = English (Scottish)
- en-us = English (US)
- en-wm = English (WM)
- eo = Esperanto
- es = Spanish
- es-la = Spanish (Latin America)
- fi = Finnish
- fr = French
- hu = Hungarian
- it = Italian
- kn = Kannada
- la = Latin
- lv = Latvian
- nl = Dutch
- pl = Polish
- pt = Portuguese (Brazil)
- pt-pt = Portuguese, European
- ro = Romanian
- sk = Slovak
- sv = Swedish
- tr = Turkish
- zh = Chinese (Mandarin)
- zh-yue = Chinese (Cantonese)
Set the desired language while instantiating the Say object:
let voice = new Say({
profile: "Cylon",
language: "es"
});
voice.say("hola mi nombre es Cylon");
Or switch languages at any time:
voice.setLanguage("es");
Or set the language as a say() option:
voice.say("bonjour je m'appelle Cylon", {
language: "fr"
});
import {MonauralScope} from 'jaxcore-say';
Include a canvas element somewhere in your page:
<canvas id="scope" />
And use the setVisualizer() method:
var canvasElement = document.getElementById('scope');
var monoScope = new MonauralScope(canvasElement);
voice.setVisualizer(monoScope);
Clone this repo, then:
cd examples/basic-example
npm install
npm start
Clone this repo, then:
npm install
npm run build-all
cd docs
python -m SimpleHTTPServer
// load http://localhost:8000
The Say Queue allows multiple Say instances (voices) to be queued up and played one by one and receive start/stop events for UI updates while that voice is speaking.
See the multiplevoices for an example of this in action.
Jaxcore Say is free software released under the GPL License.
However, IANAL (I am not a lawyer) and due to the bizarre combination of dependencies and the typical use case of compilation/minification into a JavaScript file delivered over the web, the license restrictions are ambiguous.
meSpeak (NPM module by Mikola Lysenko): https://github.com/mikolalysenko/mespeak
meSpeak (emscripten port by Norbert Landsteiner): https://www.masswerk.at/mespeak/
eSpeak http://espeak.sourceforge.net/
SAM (reverse-engineered version of SAM by Sebastian Macke) https://github.com/s-macke/SAM
SAM fork by Vidar Hokstad https://github.com/vidarh/SAM
SAM-js port by Christian Schiffler https://github.com/discordier/sam
Jaxcore Say includes modified source code from meSpeak which is GPL and also includes emscripten-compiled eSpeak code which is also GPL. Therefore this derivative work is available under the GPL.
sam-js is used as an external dependency (via NPM) for the SAM based voice profiles. sam-js was reverse engineered and could be classified as abadonware (quoted from https://github.com/s-macke/SAM) :
The software is a reverse-engineered version of a software published more than 34 years ago by "Don't ask Software".
The company no longer exists. Any attempt to contact the original authors failed. Hence S.A.M. can be best described as Abandonware (http://en.wikipedia.org/wiki/Abandonware)
As long this is the case I cannot put my code under any specific open source software license. However the software might be used under the "Fair Use" act (https://en.wikipedia.org/wiki/FAIR_USE_Act) in the USA.
Jaxcore Say will be updated according to any new information that comes to light and it is recommended that any further derivative works/improvement also be independently released under the GPL.