Implement Google network speech synthesis

chrome/browser/resources/network_speech_synthesis/tts_extension.js

Index: chrome/browser/resources/network_speech_synthesis/tts_extension.js
diff --git a/chrome/browser/resources/network_speech_synthesis/tts_extension.js b/chrome/browser/resources/network_speech_synthesis/tts_extension.js
new file mode 100644
index 0000000000000000000000000000000000000000..ebc9523f0d223ecd4ba9ed4c9b360d313f7e1875
--- /dev/null
+++ b/chrome/browser/resources/network_speech_synthesis/tts_extension.js
@@ -0,0 +1,248 @@
+// 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.

[Evan Stade, 2013/10/22 22:32:13]

I don't see where TTS is defined in this file. Explanation of that
initialization would be helpful imo.

[dmazzoni, 2013/10/22 22:45:31]

Clarified the acronym here, thanks.

+ *
+ * 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.
+ * @constructor
+ */
+function TtsExtension() {

[Evan Stade, 2013/10/22 22:32:13]

nit: {} on one line

[dmazzoni, 2013/10/22 22:45:31]

Done.

+}
+
+TtsExtension.prototype = {
+  /**
+   * The url prefix of the speech server, including static query
+   * parameters that don't change.
+   * @type {string}
+   * @const
+   * @private
+   */
+  SPEECH_SERVER_URL_:
+      'https://www.google.com/speech-api/v2/synthesize?enc=mpeg&client=chromium',

[Evan Stade, 2013/10/22 22:32:13]

80

[dmazzoni, 2013/10/22 22:45:31]

Done.

+
+  /**
+   * A mapping from language and gender to voice name, hardcoded for now
+   * until the speech synthesis server capabilities response provides this.
+   * The key of this map is of the form '<lang>-<gender>'.
+   * @type {Object.<string, string>}
+   * @private
+   */
+  LANG_AND_GENDER_TO_VOICE_NAME_: {
+    'en-gb-male': 'rjs',
+    'en-gb-female': 'fis',
+  },
+
+  /**
+   * The arguments passed to the onSpeak event handler for the utterance
+   * that's currently being spoken. Should be null when no object is
+   * pending.
+   *
+   * @type {?{utterance: string, options: Object, callback: Function}}
+   * @private
+   */
+  currentUtterance_: null,
+
+  /**
+   * The HTML5 audio element we use for playing the sound served by the
+   * speech server.
+   * @type {HTMLAudioElement}
+   * @private
+   */
+  audioElement_: null,
+
+  /**
+   * A mapping from voice name to language and gender, derived from the
+   * manifest file.  This is used in case the speech synthesis request
+   * specifies a voice name but doesn't specify a language code or gender.
+   * @type {Object.<string, {lang: string, gender: string}>}
+   * @private
+   */
+  voiceNameToLangAndGender_: {},
+
+  /**
+   * This is the main function called to initialize this extension.
+   * Initializes data structures and adds event listeners.
+   */
+  init: function() {
+    // Get voices from manifest.
+    var voices = chrome.app.getDetails().tts_engine.voices;
+    for (var i = 0; i < voices.length; i++) {
+      this.voiceNameToLangAndGender_[voices[i].voice_name] = {
+        lang: voices[i].lang,
+        gender: voices[i].gender
+      };
+    }
+
+    // Initialize the audio element and event listeners on it.
+    this.audioElement_ = document.createElement('audio');
+    document.body.appendChild(this.audioElement_);
+    this.audioElement_.addEventListener(
+        'ended', this.onStop_.bind(this), false);
+    this.audioElement_.addEventListener(
+        'canplaythrough', this.onStart_.bind(this), false);
+
+    // Install event listeners for the ttsEngine API.
+    chrome.ttsEngine.onSpeak.addListener(this.onSpeak_.bind(this));
+    chrome.ttsEngine.onStop.addListener(this.onStop_.bind(this));
+    chrome.ttsEngine.onPause.addListener(this.onPause_.bind(this));
+    chrome.ttsEngine.onResume.addListener(this.onResume_.bind(this));
+  },
+
+  /**
+   * Handler for the chrome.ttsEngine.onSpeak interface.
+   * Gets Chrome's Google API key and then uses it to generate a request
+   * url for the requested speech utterance. Sets that url as the source
+   * of the HTML5 audio element.
+   * @param {string} utterance The text to be spoken.
+   * @param {Object} options Options to control the speech, as defined
+   *     in the Chrome ttsEngine extension API.
+   * @private
+   */
+  onSpeak_: function(utterance, options, callback) {
+    // Truncate the utterance if it's too long. Both Chrome's tts
+    // extension api and the web speech api specify 32k as the
+    // maximum limit for an utterance.
+    if (utterance.length > 32768)
+      utterance = utterance.substr(0, 32768);
+
+    try {
+      // Firsttop any pending audio.

[Evan Stade, 2013/10/22 22:32:13]

First stop?

[dmazzoni, 2013/10/22 22:45:31]

Done.

+      this.onStop_();
+
+      this.currentUtterance_ = {
+        utterance: utterance,
+        options: options,
+        callback: callback
+      };
+
+      var lang = options.lang;
+      var gender = options.gender;
+      if (options.voiceName) {
+        lang = this.voiceNameToLangAndGender_[options.voiceName].lang;
+        gender = this.voiceNameToLangAndGender_[options.voiceName].gender;
+      }
+
+      // Look up the specific voice name for this language and gender.
+      // If it's not in the map, it doesn't matter - the language will
+      // be used directly. This is only used for languages where more
+      // than one gender is actually available.
+      var key = lang.toLowerCase() + '-' + gender;
+      var voiceName = this.LANG_AND_GENDER_TO_VOICE_NAME_[key];
+
+      var url = this.SPEECH_SERVER_URL_;
+      chrome.systemPrivate.getApiKey((function(key) {
+        url += '&key=' + key;
+        url += '&text=' + escape(utterance);
+        url += '&lang=' + lang.toLowerCase();
+        if (voiceName)

[Evan Stade, 2013/10/22 22:32:13]

nit: more vertical whitespace somewhere in this vicinity

[dmazzoni, 2013/10/22 22:45:31]

Done.

+          url += '&name=' + voiceName;
+        if (options.rate) {
+          // Input rate is between 0.1 and 10.0 with a default of 1.0.
+          // Output speed is between 0.0 and 1.0 with a default of 0.5.
+          url += '&speed=' + (options.rate / 2.0);
+        }
+        if (options.pitch) {
+          // Input pitch is between 0.0 and 2.0 with a default of 1.0.
+          // Output pitch is between 0.0 and 1.0 with a default of 0.5.
+          url += '&pitch=' + (options.pitch / 2.0);
+        }
+
+        // This begins loading the audio but does not play it.
+        // When enough of the audio has loaded to begin playback,
+        // the 'canplaythrough' handler will call this.onStart_,
+        // which sends a start event to the ttsEngine callback and
+        // then begins playing audio.
+        this.audioElement_.src = url;
+      }).bind(this));
+    } catch (err) {
+      console.error(String(err));
+      callback({
+        'type': 'error',
+        'errorMessage': String(err)
+      });
+      this.currentUtterance_ = null;
+    }
+  },
+
+  /**
+   * Handler for the chrome.ttsEngine.onStop interface.
+   * Called either when the ttsEngine API requests us to stop, or when
+   * we reach the end of the audio stream. Pause the audio element to
+   * silence it, and send a callback to the ttsEngine API to let it know
+   * that we've completed. Note that the ttsEngine API manages callback
+   * messages and will automatically replace the 'end' event with a
+   * more specific callback like 'interrupted' when sending it to the
+   * TTS client.
+   * @private
+   */
+  onStop_: function() {
+    if (this.currentUtterance_) {
+      this.audioElement_.pause();
+      this.currentUtterance_.callback({
+        'type': 'end',
+        'charIndex': this.currentUtterance_.utterance.length
+      });
+    }
+    this.currentUtterance_ = null;
+  },
+
+  /**
+   * Handler for the canplaythrough event on the audio element.
+   * Called when the audio element has buffered enough audio to begin
+   * playback. Send the 'start' event to the ttsEngine callback and
+   * then begin playing the audio element.
+   * @private
+   */
+  onStart_: function() {
+    if (this.currentUtterance_) {
+      if (this.currentUtterance_.options.volume !== undefined) {
+        // Both APIs use the same range for volume, between 0.0 and 1.0.
+        this.audioElement_.volume = this.currentUtterance_.options.volume;
+      }
+      this.audioElement_.play();
+      this.currentUtterance_.callback({
+          'type': 'start',
+          'charIndex': 0
+      });
+    }
+  },
+
+  /**
+   * Handler for the chrome.ttsEngine.onPause interface.
+   * Pauses audio if we're in the middle of an utterance.
+   * @private
+   */
+  onPause_: function() {
+    if (this.currentUtterance_) {
+      this.audioElement_.pause();
+    }
+  },
+
+  /**
+   * Handler for the chrome.ttsEngine.onPause interface.
+   * Resumes audio if we're in the middle of an utterance.
+   * @private
+   */
+  onResume_: function() {
+    if (this.currentUtterance_) {
+      this.audioElement_.play();
+    }
+  }
+
+};
+
+(new TtsExtension()).init();