Update app.py

app.py

@@ -50,19 +50,29 @@ def transcribe(audio_file, task):
 
     # Read audio using ffmpeg_read (correcting input format)
     audio_array = ffmpeg_read(audio_data, pipe.feature_extractor.sampling_rate)
+    
+    duration = len(audio_array) / pipe.feature_extractor.sampling_rate
+    print(f"Audio duration: {duration:.2f} seconds")
 
     # Convert to proper format
     inputs = {
-        "raw": np.array(audio_array),  
+        "array": np.array(audio_array),  
         "sampling_rate": pipe.feature_extractor.sampling_rate
     }
-
+    
+    generate_kwargs = {
+        "task": task,
+        "no_speech_threshold": 0.3,
+        "logprob_threshold": -1.0,
+        "compression_ratio_threshold": 2.4
+    }
+    
     # Perform transcription
     result = pipe(
         inputs,
         batch_size=BATCH_SIZE,
-        generate_kwargs={"task": task},
-        return_timestamps=True
+        generate_kwargs=generate_kwargs,
+        return_timestamps="word"
     )
 
     return result["text"]
@@ -143,6 +153,6 @@ yt_transcribe = gr.Interface(
 )
 
 with demo:
-    gr.TabbedInterface([file_transcribe, yt_transcribe], ["Audio file", "YouTube"])
+    gr.TabbedInterface([file_transcribe], ["Audio file"])
 
 demo.launch()