Update app.py

app.py

@@ -37,6 +37,31 @@ pipe = pipeline(
     device=device,
 )
 
+@spaces.GPU
+def stream_transcribe(stream, new_chunk):
+    start_time = time.time() 
+    try:
+        sr, y = new_chunk
+        
+        # Convert to mono if stereo
+        if y.ndim > 1:
+            y = y.mean(axis=1)
+            
+        y = y.astype(np.float32)
+        y /= np.max(np.abs(y))
+    
+        if stream is not None:
+            stream = np.concatenate([stream, y])
+        else:
+            stream = y
+
+        end_time = time.time()
+        latency = end_time - start_time
+        return stream, transcriber({"sampling_rate": sr, "raw": stream})["text"], f"{latency:.2f}"
+    except Exception as e:
+        print(f"Error during Transcription: {e}")
+        return previous_transcription, "Error"
+
 @spaces.GPU
 def transcribe(inputs, previous_transcription):
     start_time = time.time() 
@@ -86,8 +111,8 @@ with gr.Blocks() as microphone:
             latency_textbox = gr.Textbox(label="Latency (seconds)", value="0.0", scale=0)
         with gr.Row():
             clear_button = gr.Button("Clear Output")
-
-        input_audio_microphone.stream(transcribe, [input_audio_microphone, output], [output, latency_textbox], time_limit=45, stream_every=2, concurrency_limit=None)
+        state = gr.State()
+        input_audio_microphone.stream(stream_transcribe, [state, input_audio_microphone], [state, output, latency_textbox], time_limit=45, stream_every=1, concurrency_limit=None)
         clear_button.click(clear, outputs=[output])
 
 with gr.Blocks() as file: