Update app.py

app.py

@@ -479,17 +479,24 @@ base_audio_drive = "/data/audio"
 #     return stream, full_text, result
 
 
+# def transcribe_function(stream, new_chunk):
+#     sr, y = new_chunk[0], new_chunk[1]
+#     y = y.astype(np.float32) / np.max(np.abs(y))
+#     if stream is not None:
+#         stream = np.concatenate([stream, y])
+#     else:
+#         stream = y
+#     result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
+#     full_text = result.get("text", "")
+#     return stream, full_text  # Return the transcribed text
+
 def transcribe_function(stream, new_chunk):
     sr, y = new_chunk[0], new_chunk[1]
     y = y.astype(np.float32) / np.max(np.abs(y))
-    if stream is not None:
-        stream = np.concatenate([stream, y])
-    else:
-        stream = y
+    stream = y  # Always start with fresh stream
     result = pipe_asr({"array": stream, "sampling_rate": sr}, return_timestamps=False)
     full_text = result.get("text", "")
-    return stream, full_text  # Return the transcribed text
-
+    return stream, full_text
 
 
 def update_map_with_response(history):
@@ -703,7 +710,8 @@ with gr.Blocks(theme='Pijush2023/scikit-learn-pijush') as demo:
             clear_button.click(fn=clear_textbox, inputs=None, outputs=chat_input)
            
             audio_input = gr.Audio(sources=["microphone"], streaming=True, type='numpy')
-            audio_input.stream(transcribe_function, inputs=[state, audio_input], outputs=[state, chat_input], api_name="voice_query_to_text")    
+            # audio_input.stream(transcribe_function, inputs=[state, audio_input], outputs=[state, chat_input], api_name="voice_query_to_text")    
+            audio_input.stream(transcribe_function, inputs=[None, audio_input], outputs=[None, chat_input], api_name="voice_query_to_text")