update streaming audio generation

stream_app.py

@@ -63,25 +63,40 @@ def run_gradio(config:dict):
         return input_text
 
     # reset transcribed text
-    def audio_reset(input_text):
+    def audio_reset(input_text, path_prior):
+        if path_prior is not None:
+            Path(path_prior).unlink()
         # audio = whisper.clear?
-        return ""    # return empty
+        return "", None    # return empty, clear prior file
     
     # speak input text
-    def audio_speak(input_text, speaker_name, output_complete, auto_speak=None):
-        print(f"OUTPUT: {output_complete}")
-        if not output_complete:
-            return None
-        if auto_speak is not None and auto_speak.lower() == "manual":   # abort if manual
-            return None
-        temp_file = tempfile.NamedTemporaryFile(delete=False)
+    def audio_speak(input_text, speaker_name, input_done=True, offset_prior=0, path_prior=None, auto_speak=None):
+        # alternate on-device? - https://github.com/suno-ai/bark?tab=readme-ov-file
+        # print(f"Speak: {input_text}, {offset_prior} of {len(input_text)}")
+        if not input_text:                          # empty string on conclusion (when streaming)
+            return None, None, 0
+        elif auto_speak is not None:
+            if "manual" in auto_speak.lower():      # don't proceed if manual
+                return None, None, 0
+            elif (not input_done) and ("stream" not in auto_speak.lower()):    # stream, not done
+                return None, None, 0
+        
+        if (path_prior is None) or (offset_prior > len(input_text)):
+            temp_file = tempfile.NamedTemporaryFile(delete=False)
+            path_prior = temp_file.name
+            offset_prior = 0
         response = client.audio.speech.create(
             model="tts-1",
             voice=speaker_name,
-            input=input_text
+            input=input_text[offset_prior:]
         )
-        response.write_to_file(temp_file.name)
-        return temp_file.name
+        offset_prior += len(input_text)
+        # append to existing file 
+        #   example: https://community.openai.com/t/streaming-from-text-to-speech-api/493784/5
+        with open(path_prior, 'ab') as file_append:
+            for chunk in response.iter_bytes(chunk_size=4096):
+                file_append.write(chunk)
+        return path_prior, path_prior, offset_prior
 
 
     # Define Gradio interface
@@ -102,12 +117,12 @@ def run_gradio(config:dict):
         )
 
         partial_response = ""
-        for stream_response in response:
-            logger.warning(f"Prompt response: {stream_response.to_dict()}")
-            token = stream_response.choices[0].delta.content
-            if token is None:
+        response_dicts = [stream_response.to_dict() for stream_response in response]
+        logger.warning(f"Prompt response: {response_dicts}")
+        for stream_response in response_dicts:
+            if 'content' not in stream_response['choices'][0]['delta']:
                 break
-            partial_response += token
+            partial_response += stream_response['choices'][0]['delta']['content']
             yield partial_response, False
         yield partial_response, True
 
@@ -161,41 +176,47 @@ def run_gradio(config:dict):
                             choices=["alloy", "echo", "fable", "onyx", "nova", "shimmer"], 
                             show_label=False, value="nova", interactive=True,
                         )
+                    with gr.Row():
                         combo_autospeak = gr.Radio(
-                            choices=["Auto-speak", "Manual"], show_label=False,
+                            choices=["Auto-speak", "Auto-speak (stream)", "Manual"], show_label=False,
                             value="Manual", interactive=True,
                         )
                     with gr.Row():
                         speak_button = gr.Button("Speak!", variant='secondary', interactive=True)
                     with gr.Row():
                         audio_playback = gr.Audio(
-                            label="Speech", autoplay=True,
-                            streaming=False,
+                            label="Speech", autoplay=True, streaming=False,
                             type="filepath", sources=None,
                         )
 
         with gr.Row():
-            submit_button = gr.Button("Submit", variant='primary')
-            output_complete = gr.State(False)
-                            
+            submit_button = gr.Button("Generate Response", variant='primary')
+            generate_done = gr.State(False)             # is last genai content chunked?
+            path_prior = gr.State(None)                 # retain prior file for audio playback
+            offset_prior = gr.State(0)                  # track textual offset in genrated content
+
                 
-        audio_input.stream(audio_transcribe,
+        audio_input.stream(audio_transcribe,            # started streaming to transcribe
             inputs=[audio_input_model, audio_input, audio_threshold, input_text],
             outputs=input_text)
-        audio_input.clear(audio_reset, inputs=input_text, outputs=input_text)
-        audio_input.start_recording(audio_reset, inputs=input_text, outputs=input_text)
-        audio_input.stop_recording(get_ai_response, 
+        audio_input.clear(audio_reset,                  # cleared audio
+            inputs=[input_text, path_prior], 
+            outputs=[input_text, path_prior])
+        audio_input.start_recording(audio_reset,        # started a new speech->text
+            inputs=[input_text, path_prior], 
+            outputs=[input_text, path_prior])
+        audio_input.stop_recording(get_ai_response,     # stopped recording, start response
             inputs=[input_text], 
-            outputs=[output_text, output_complete])
-        submit_button.click(get_ai_response, 
+            outputs=[output_text, generate_done])
+        submit_button.click(get_ai_response,            # clicked 'generate'
             inputs=[input_text], 
-            outputs=[output_text, output_complete])
-        output_text.change(audio_speak, 
-            inputs=[output_text, combo_speaker, output_complete, combo_autospeak],
-            outputs=audio_playback)
-        speak_button.click(audio_speak, 
-            inputs=[output_text, combo_speaker, output_complete],
-            outputs=audio_playback)
+            outputs=[output_text, generate_done])
+        output_text.change(audio_speak,               # streaming response from generate
+            inputs=[output_text, combo_speaker, generate_done, offset_prior, path_prior, combo_autospeak],
+            outputs=[audio_playback, path_prior, offset_prior])
+        speak_button.click(audio_speak,                 # click for speak trigger
+            inputs=[output_text, combo_speaker],
+            outputs=[audio_playback, path_prior, offset_prior])
         
 
     # demo.set_api_mode(enabled=False)   # Disable API exposure
@@ -237,7 +258,6 @@ def parse_args() -> dict:
 if __name__ == "__main__":
     os.environ['GRADIO_ANALYTICS_ENABLED'] = 'False'
     api_key = os.environ.get("OPENAI_API_KEY")
-    print(api_key)
     if not api_key:
         raise ValueError("OPENAI_API_KEY environment variable not set as environment variable or as a setting in `.env`.  (see https://platform.openai.com/docs/quickstart/step-2-set-up-your-api-key)")