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..d4e74295fbfc8cd63fa6b986cc28f225f61d271a
--- /dev/null
+++ b/chrome/browser/resources/network_speech_synthesis/tts_extension.js
@@ -0,0 +1,101 @@
+// 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() {
+}
+
+TtsExtension.prototype = {
+  speechServerUrl_:
+      'https://www.google.com/speech-api/v2/synthesize?enc=mpeg&client=chrome',

[gshires1, 2013/10/17 00:34:50]

ASR uses client=chromium
(Unless there's other reasoning or precedents, it probably makes sense to be
consistent and use chromium.)

[dmazzoni, 2013/10/17 17:14:00]

Done.

+
+  currentUtterance_: null,
+
+  audioElement_: null,
+
+  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) {
+    try {
+      this.onStop_();
+      this.currentUtterance_ = {
+        utterance: utterance,
+        options: options,
+        callback: callback
+      };
+
+      var url = this.speechServerUrl_;
+      chrome.systemPrivate.getApiKey((function(key) {
+        url += '&key=' + key;
+        url += '&text=' + escape(utterance);

[gshires1, 2013/10/17 00:34:50]

How long might text be?  (Assuming it could be quite long, might that cause a
problem with a long URL?)
I think it's better to send text as POST data.  (Also, POST might be better in
terms of not having to escape it.)

[dmazzoni, 2013/10/17 17:14:00]

The maximum length is a good point, but I still prefer GET.

As a quick fix, I truncated the string to 32768 characters after escaping.
That's the same maximum we enforce in the TTS extension API already. Chrome can
handle much longer http urls, but really the client ought to be splitting it up
into smaller pieces - that's a lot of text already.

Switching to post would be nontrivial and may not be ideal. Right now we're
setting this url as the source of an html5 <audio> element; this gives us
streaming and caching automatically.

If we wanted to use POST, we'd have to use XmlHttpRequest and then send the
results to the web audio api. That would mean we'd have to wait for the full
audio file to finish downloading first. It also wouldn't be allowed to cache the
audio data in this case.

+        url += '&lang=' + options.lang.toLowerCase();
+        if (options.rate) {
+          url += '&speed=' + (options.rate / 2.0);
+        }
+        if (options.pitch) {
+          url += '&pitch=' + (options.pitch / 2.0);
+        }
+        if (options.volume) {
+          url += '&vol=' + options.volume;
+        }
+        this.audioElement_.src = url;
+      }).bind(this));
+    } catch (err) {
+      console.log(String(err));
+      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();