Add --use-text-input and --use-text-output.

Add --use-text-input and --use-text-output.

Author: patrickxia

google-assistant-sdk/googlesamples/assistant/grpc/pushtotalk.py

@@ -41,13 +41,15 @@
         assistant_helpers,
         audio_helpers,
         browser_helpers,
-        device_helpers
+        device_helpers,
+        text_helpers
     )
 except (SystemError, ImportError):
     import assistant_helpers
     import audio_helpers
     import browser_helpers
     import device_helpers
+    import text_helpers
 
 
 ASSISTANT_API_ENDPOINT = 'embeddedassistant.googleapis.com'
@@ -256,6 +258,10 @@ def gen_assist_requests(self):
               help='Enable visual display of Assistant responses in HTML.')
 @click.option('--verbose', '-v', is_flag=True, default=False,
               help='Verbose logging.')
+@click.option('--use-text-input', is_flag=True, default=False,
+              help='Use the hacky text-to-speech stuff.')
+@click.option('--use-text-output', is_flag=True, default=False,
+              help='Use the hacky speech-to-text stuff.')
 @click.option('--input-audio-file', '-i',
               metavar='<input file>',
               help='Path to input audio file. '
@@ -293,7 +299,8 @@ def gen_assist_requests(self):
               help='Force termination after a single conversation.')
 def main(api_endpoint, credentials, project_id,
          device_model_id, device_id, device_config,
-         lang, display, verbose,
+         lang, display, verbose, use_text_input,
+         use_text_output,
          input_audio_file, output_audio_file,
          audio_sample_rate, audio_sample_width,
          audio_iter_size, audio_block_size, audio_flush_size,
@@ -336,7 +343,13 @@ def main(api_endpoint, credentials, project_id,
 
     # Configure audio source and sink.
     audio_device = None
-    if input_audio_file:
+    if use_text_input:
+        audio_source = text_helpers.TextSource(
+            "",
+            sample_rate=audio_sample_rate,
+            sample_width=audio_sample_width
+        )
+    elif input_audio_file:
         audio_source = audio_helpers.WaveSource(
             open(input_audio_file, 'rb'),
             sample_rate=audio_sample_rate,
@@ -351,7 +364,13 @@ def main(api_endpoint, credentials, project_id,
                 flush_size=audio_flush_size
             )
         )
-    if output_audio_file:
+
+    if use_text_output:
+        audio_sink = text_helpers.TextSink(
+            sample_rate=audio_sample_rate,
+            sample_width=audio_sample_width
+        )
+    elif output_audio_file:
         audio_sink = audio_helpers.WaveSink(
             open(output_audio_file, 'wb'),
             sample_rate=audio_sample_rate,
@@ -452,8 +471,15 @@ def blink(speed, number):
         # and playing back assistant response using the speaker.
         # When the once flag is set, don't wait for a trigger. Otherwise, wait.
         wait_for_user_trigger = not once
+        first = True
         while True:
-            if wait_for_user_trigger:
+            if use_text_input:
+                if not first and use_text_output:
+                    logging.info('Recognized response from Assistant: ' + audio_sink.recognize())
+                    audio_sink.reset()
+                audio_source.reset_text(click.prompt(''))
+                first = False
+            elif wait_for_user_trigger:
                 click.pause(info='Press Enter to send a new request...')
             continue_conversation = assistant.assist()
             # wait for user trigger if there is no follow-up turn in

google-assistant-sdk/googlesamples/assistant/grpc/text_helpers.py

@@ -0,0 +1,67 @@
+import io
+import wave
+
+try:
+    from . import audio_helpers
+except (SystemError, ImportError):
+    # needs ffmpeg
+    import audio_helpers
+
+import speech_recognition as sr
+
+from gtts import gTTS
+from pydub import AudioSegment
+
+class TextSource(audio_helpers.WaveSource):
+    """Audio source that uses text to speech as the
+    backing store.
+
+    Silence is returned when there is no current text.
+    Args:
+      text: the initial text to read.
+      sample_rate: sample rate in hertz.
+      sample_width: size of a single sample in bytes.
+    """
+    def __init__(self, text, sample_rate, sample_width):
+        self._text = text
+        self._sample_rate = sample_rate
+        self._sample_width = sample_width
+        self._sleep_until = 0
+        self._wavep = None
+
+        if text:
+            self.reset_text(text)
+
+    def reset_text(self, text):
+        self._text = text
+        with io.BytesIO() as fp:
+            tts = gTTS(text, 'en')
+            tts.write_to_fp(fp)
+            fp.seek(0)
+            data = AudioSegment.from_mp3(fp)._data
+            self._fp = io.BytesIO(data)
+
+    def close(self):
+        """Close the underlying stream."""
+        pass
+
+class TextSink(audio_helpers.WaveSink):
+    def __init__(self, sample_rate, sample_width):
+        self._sample_rate = sample_rate
+        self._sample_width = sample_width
+
+        self._r = sr.Recognizer()
+        self.reset()
+
+    def recognize(self):
+        self._buf.seek(0)
+        with sr.AudioFile(self._buf) as source:
+            audio = self._r.record(source)
+        text = self._r.recognize_google(audio)
+        return text
+
+    def reset(self):
+        self._buf = io.BytesIO()
+        super().__init__(self._buf, self._sample_rate, self._sample_width)
+
+