git add requirements.txt git commit -m "Install openai-whisper instead of whisper" git push

app.py

@@ -7,24 +7,18 @@ import gradio as gr
 # 1) Load your balanced text classifier
 text_clf = joblib.load("text_pipeline_balanced.joblib")
 
-# 2) Load Whisper-Large-v2 via openai/whisper on CPU
-model = whisper.load_model("large-v2")  # Change to "base" if you want smaller
+# 2) Load Whisper-Large-v2 via official OpenAI Whisper on CPU
+model = whisper.load_model("large-v2")  # or "base" for a smaller model
 
 def classify(audio_path):
     """
     audio_path: str → path to the uploaded file
     returns: transcript (str), safety probabilities (dict), unsafe probability (str)
     """
-    # Read & convert to mono 16k WAV
-    audio, sr = sf.read(audio_path, dtype="float32")
-    if audio.ndim > 1:
-        audio = audio.mean(axis=1)
-    # Whisper’s .transcribe will resample internally if needed
-
     # Run beam search transcription
     result = model.transcribe(
         audio_path,
-        beam_size=5,         # beam search for quality
+        beam_size=5,         # beam search for higher accuracy
         language="en"
     )
     txt = result["text"].strip()
@@ -47,7 +41,7 @@ iface = gr.Interface(
     inputs=audio_input,
     outputs=[transcript_out, probs_out, unsafe_out],
     title="BubbleGuard Audio Safety Checker",
-    description="Uses the official whisper package for identical, CPU-only transcripts."
+    description="Uses the official openai-whisper package for identical, CPU-only transcripts."
 )
 
 if __name__ == "__main__":