audio_processing.py

import whisper
import os

# --- CONFIGURATION ---
# "base" is a good balance of speed and accuracy. 
# Use "tiny" for super speed, or "small" for better accuracy.
WHISPER_MODEL_SIZE = "base" 

class AudioAnalyzer:
    def __init__(self):
        print(f"Loading Whisper AI ({WHISPER_MODEL_SIZE})...")
        self.model = whisper.load_model(WHISPER_MODEL_SIZE)
        print("Whisper AI loaded successfully.")

    def process_audio(self, audio_path):
        """
        Takes an audio file path, converts to text, and extracts basic features.
        """
        if not os.path.exists(audio_path):
            return {"error": "File not found"}

        print(f"Transcribing {audio_path}...")
        
        # 1. TRANSCRIBE (Speech -> Text)
        result = self.model.transcribe(audio_path, language="en", fp16=False)
        transcribed_text = result["text"].strip()
        
        # 2. EXTRACT META-FEATURES (Bonus Signals)
        # We can detect 'slow speech' (Psychomotor retardation) common in Depression
        duration_seconds = result['segments'][-1]['end'] if result['segments'] else 0
        word_count = len(transcribed_text.split())
        
        wpm = (word_count / duration_seconds) * 60 if duration_seconds > 0 else 0
        
        # Interpret Rate of Speech
        speech_rate_label = "Normal"
        if wpm < 110:
            speech_rate_label = "Slow (Possible Depression Indicator)"
        elif wpm > 160:
            speech_rate_label = "Fast (Possible Anxiety/ADHD Indicator)"

        return {
            "text": transcribed_text,
            "wpm": round(wpm, 1),
            "rate_label": speech_rate_label
        }

# --- TEST IT ---
if __name__ == "__main__":
    # You need a dummy audio file to test this. 
    # If on Colab, upload a file named 'test_audio.mp3'
    
    analyzer = AudioAnalyzer()
    
    # Example usage (Uncomment if you have a file)
    # output = analyzer.process_audio("test_audio.mp3")
    # print(output)