Stream speech audio to server as it gets recorded, instead of waiting until end of recording.

content/browser/speech/speech_recognition_request.h

Index: content/browser/speech/speech_recognition_request.h
diff --git a/content/browser/speech/speech_recognition_request.h b/content/browser/speech/speech_recognition_request.h
index 3036d59bb8da5e4b00ab46db33a23caaafcf7b85..3bb629869daf695e5611d5152a136a0c3a7ae31f 100644
--- a/content/browser/speech/speech_recognition_request.h
+++ b/content/browser/speech/speech_recognition_request.h
@@ -46,12 +46,14 @@ class SpeechRecognitionRequest : public URLFetcher::Delegate {
   // Sends a new request with the given audio data, returns true if successful.
   // The same object can be used to send multiple requests but only after the
   // previous request has completed.
-  bool Send(const std::string& language,
-            const std::string& grammar,
-            const std::string& hardware_info,
-            const std::string& origin_url,
-            const std::string& content_type,
-            const std::string& audio_data);
+  void Start(const std::string& language,
+             const std::string& grammar,
+             const std::string& hardware_info,
+             const std::string& origin_url,
+             const std::string& content_type);
+
+  // Send a single chunk of audio immediately to the server.
+  void UploadAudioChunk(const std::string& audio_data, bool is_last_chunk);
 
   bool HasPendingRequest() { return url_fetcher_ != NULL; }