Add check if audio is long

multimodality_tools.py

@@ -27,7 +27,10 @@ def transcribe_audio(file_id: str) -> str:
 
     asr = pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
 
-    output = asr(audio)
+    if (len(audio) / 16000) > 25:
+        output = asr(audio, return_timestamps=True)
+    else:
+        output = asr(audio)
 
     return output["text"].strip()