Update app.py

app.py

@@ -1,10 +1,10 @@
 import gradio as gr
-from transformers import pipeline
-from pydub import AudioSegment
+import whisper
 import os
+from pydub import AudioSegment
 
-# Load a smaller Whisper model for faster transcription
-model = pipeline("automatic-speech-recognition", model="openai/whisper-base")
+# Load the Whisper model
+model = whisper.load_model("base")  # Use "base" for faster processing
 
 def split_audio(filepath, chunk_length_ms=30000):
     """Split audio into chunks of `chunk_length_ms` milliseconds."""
@@ -26,12 +26,12 @@ def transcribe_audio(audio_file):
     detected_language = None
     
     for chunk in chunks:
-        # Enable language detection and transcription
-        result = model(chunk, generate_kwargs={"task": "transcribe", "language": None})  # Let Whisper detect language
+        # Transcribe the chunk and detect the language
+        result = model.transcribe(chunk, fp16=False)  # Set fp16=False if not using GPU
         transcriptions.append(result["text"])
         
-        # Extract detected language from the result (if available)
-        if "language" in result:
+        # Extract detected language from the result
+        if detected_language is None and "language" in result:
             detected_language = result["language"]
         
         os.remove(chunk)  # Clean up chunk files