Update app.py

app.py

@@ -4,6 +4,8 @@ import torch
 import librosa
 import json
 
+max_duration = int(30 * sr)
+
 def load_model(model_name = "cawoylel/windanam_mms-1b-tts_v2"):
   """
   Function to load model from hugging face
@@ -32,7 +34,9 @@ def transcribe(audio_file_mic=None, audio_file_upload=None):
     speech, sample_rate = librosa.load(audio_file)
     if sample_rate != 16000:
         speech = librosa.resample(speech, orig_sr=sample_rate, target_sr=16000)
-
+    duration = librosa.get_duration(speech)
+    if duration > 30:
+        speech = speech[:max_duration]
     return transcribe_audio(speech)