Update app.py

app.py

@@ -1,8 +1,14 @@
 import gradio as gr
 from transformers import pipeline
+from faster_whisper import WhisperModel
+import librosa
 
 # Load the model
-pipe = pipeline("automatic-speech-recognition", model="vargha/whisper-large-v3")
+# pipe = pipeline("automatic-speech-recognition", model="navidved/persian-whisper-large-v3-ct2")
+model = WhisperModel("navidved/persian-whisper-large-v3-ct2", device="cpu", compute_type="int8")
+
+# Define the maximum audio length in seconds
+MAX_AUDIO_LENGTH = 40  # 40 seconds
 
 # Define the inference function
 def transcribe_audio(audio):
@@ -10,8 +16,16 @@ def transcribe_audio(audio):
         return "No audio file uploaded. Please try again."
     
     try:
+        audio_data, sr = librosa.load(audio, sr=None)
+        duration = librosa.get_duration(y=audio_data, sr=sr)
+
+        # Check if the audio is longer than the allowed duration
+        if duration > MAX_AUDIO_LENGTH:
+            return f"Audio is too long. Please upload an audio file shorter than {MAX_AUDIO_LENGTH} seconds."
+        
         # Perform transcription
-        result = pipe(audio)["text"]
+        segments, _ = model.transcribe(audio, vad_filter=True)
+        result = segments.text
         return result
     except Exception as e:
         return f"Error during transcription: {str(e)}"