Update app.py

app.py

@@ -156,7 +156,16 @@ def recognize_speech(audio_path):
         inputs = {k: v.to(device) for k, v in inputs.items()}
 
         # Generate transcription
-        generated_ids = speech_recognizer.generate(**inputs)
+        #generated_ids = speech_recognizer.generate(**inputs)
+            generated_ids = speech_recognizer.generate(
+        **inputs,
+        max_length=100,
+        num_beams=3,  # Reduced from default
+        early_stopping=False,  # Disabled
+        no_repeat_ngram_size=3,
+        temperature=0.7,  # Added for stability
+        do_sample=False  # Disable sampling
+            )
         transcription = speech_processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
 
         return transcription