Implement Google network speech synthesis

chrome/browser/resources/network_speech_synthesis/tts_extension.js

+// Copyright 2013 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+/**
+ * @fileoverview
+ * This is a component extension that implements a speech engine
+ * powered by Google's speech synthesis API.
+ *
+ * This is an "event page", so it's not loaded when the API isn't being used,
+ * and doesn't waste resources. When a web page or web app makes a speech
+ * request and the parameters match one of the voices in this extension's
+ * manifest, it makes a request to Google's API using Chrome's private key
+ * and plays the resulting speech using HTML5 audio.
+ */
+
+/**
+ * The main class for this extension. Adds listeners to
+ * chrome.ttsEngine.onSpeak and chrome.ttsEngine.onStop and implements
+ * them using Google's speech synthesis API.
+ */
+function TtsExtension() {

[David Tseng, 2013/10/17 20:43:49]

Do you want to declare this function globally?
Perhaps do:
var TtsExtension = function () {};

(Also use JS doc).

[dmazzoni, 2013/10/18 16:23:19]

I think this is the Chrome idiomatic style for declaring a class, and plus
there's no reason not to make it a global since it's just an extension event
page. Added @constructor to jsdoc.

+}
+
+TtsExtension.prototype = {

[David Tseng, 2013/10/17 20:56:30]

I personally don't mind this style of js, but let's stick with the standard JS
style guide and do
TtsExtension.prototype.foo = function()

and initialize your instance members above in the constructor.

[dmazzoni, 2013/10/18 16:23:19]

I don't have a strong preference, I was just trying to copy the idiomatic style
of Chrome javascript, especially other code in chrome/browser/resources such as
ntp4 and options. If there's a specific push towards another style I'm happy to
go with that.

+  speechServerUrl_:

[David Tseng, 2013/10/17 20:43:49]

@private, @const; SPEECH_SERVER_URL_

[dmazzoni, 2013/10/18 16:23:19]

Done.

+      'https://www.google.com/speech-api/v2/synthesize?enc=mpeg&client=chromium',
+
+  currentUtterance_: null,
+
+  audioElement_: null,

[David Tseng, 2013/10/17 20:43:49]

JS doc for all of these please.

[dmazzoni, 2013/10/18 16:23:19]

Done.

+
+  run: function() {
+    this.audioElement_ = document.createElement('audio');
+    document.body.appendChild(this.audioElement_);
+    this.audioElement_.addEventListener('ended', (function() {
+      this.onStop_();
+    }).bind(this), false);
+
+    this.audioElement_.addEventListener('canplaythrough', (function() {
+      if (this.currentUtterance_) {
+        this.audioElement_.play();
+        this.currentUtterance_.callback({
+            'type': 'start',
+            'charIndex': 0
+        });
+      }
+    }).bind(this), false);
+
+    chrome.ttsEngine.onSpeak.addListener(this.onSpeak_.bind(this));
+    chrome.ttsEngine.onStop.addListener(this.onStop_.bind(this));
+  },
+
+  onSpeak_: function(utterance, options, callback) {
+    // Truncate the utterance if it's too long.

[David Tseng, 2013/10/17 20:43:49]

Why not just chunk the utterance? (and queue up additional utterances to be
requested).

[dmazzoni, 2013/10/18 16:23:19]

Both Chrome's tts extension api and the web speech api specify 32k as the
maximum limit for an utterance. I think it's better to have a limit like this -
that's quite a lot of text already - rather than worry about potential bugs or
memory leaks that could happen with unreasonably large input.

+    utterance = escape(utterance);
+    if (utterance.length > 32768)
+      utterance = utterance.substr(0, 32768);
+
+    try {
+      this.onStop_();

[David Tseng, 2013/10/17 20:43:49]

Can we get this engine to respect queue mode? (perhaps maintaining a queue of
utterances and clearing on flush).

[dmazzoni, 2013/10/18 16:23:19]

The queue is managed by Chrome. A ttsEngine just needs to handle one utterance
at a time.

+      this.currentUtterance_ = {
+        utterance: utterance,
+        options: options,
+        callback: callback
+      };
+
+      var url = this.speechServerUrl_;
+      chrome.systemPrivate.getApiKey((function(key) {
+        url += '&key=' + key;
+        url += '&text=' + utterance;
+        url += '&lang=' + options.lang.toLowerCase();
+        if (options.rate) {
+          url += '&speed=' + (options.rate / 2.0);

[David Tseng, 2013/10/17 20:43:49]

Can you comment on the range expected?

[dmazzoni, 2013/10/18 16:23:19]

Done.

+        }
+        if (options.pitch) {
+          url += '&pitch=' + (options.pitch / 2.0);

[David Tseng, 2013/10/17 20:43:49]

Ditto.

[dmazzoni, 2013/10/18 16:23:19]

Done.

+        }
+        if (options.volume) {
+          url += '&vol=' + options.volume;

[David Tseng, 2013/10/17 20:43:49]

Couldn't we just handle volume directly in the playback?

[dmazzoni, 2013/10/18 16:23:19]

Sure, done. Actually I see one potential advantage of that is that it could
cache the same utterance to play at a different volume.

+        }
+        this.audioElement_.src = url;
+      }).bind(this));
+    } catch (err) {
+      console.log(String(err));

[David Tseng, 2013/10/17 20:43:49]

console.error

[dmazzoni, 2013/10/18 16:23:19]

Done.

+      callback({
+        'type': 'error',
+        'errorMessage': String(err)
+      });
+      this.currentUtterance_ = null;
+    }
+  },
+
+  onStop_: function() {
+    if (this.currentUtterance_) {
+      this.audioElement_.pause();
+      this.currentUtterance_.callback({
+        'type': 'end',
+        'charIndex': this.currentUtterance_.utterance.length
+      });
+    }
+    this.currentUtterance_ = null;
+  }
+};
+
+(new TtsExtension()).run();