Update Whisper: use 'small' model with auto language detection

app.py

@@ -111,10 +111,10 @@ def transcribe_audio_whisper(audio_file):
 
     try:
         # Load Whisper model (will be done on GPU)
-        model = whisper.load_model("base")
+        model = whisper.load_model("small")
 
         # Transcribe the audio
-        result = model.transcribe(audio_file, language="en")
+        result = model.transcribe(audio_file)
 
         return result["text"].strip()