app.py

import gradio as gr
import numpy as np
import librosa
from transformers import pipeline

# Initialize models
emotion_analyzer = pipeline("audio-classification", model="MIT/ast-finetuned-speech-commands-v2")
speech_recognizer = pipeline("automatic-speech-recognition", 
                           model="kresnik/wav2vec2-large-xlsr-korean")

def analyze_voice(audio_file):
    """Voice analysis function"""
    try:
        # Load audio
        y, sr = librosa.load(audio_file)
        
        # 1. Voice emotion analysis
        emotions = emotion_analyzer(y)
        primary_emotion = emotions[0]
        
        # 2. Speech to text
        text_result = speech_recognizer(y)
        
        # 3. Extract audio features
        mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13)
        energy = np.mean(librosa.feature.rms(y=y))
        
        return {
            "emotion": primary_emotion['label'],
            "emotion_probability": f"{primary_emotion['score']:.2f}",
            "transcribed_text": text_result['text'],
            "energy_level": f"{energy:.2f}",
            "status": "Analysis complete"
        }
    except Exception as e:
        return {
            "error": str(e),
            "status": "Error occurred"
        }

# Create Gradio interface
interface = gr.Interface(
    fn=analyze_voice,
    inputs=gr.Audio(source="microphone", type="filepath", label="Voice Input"),
    outputs=gr.JSON(label="Analysis Results"),
    title="Digital Gut - Voice Emotion Analysis",
    description="Performs emotion analysis and text conversion from voice input.",
    theme=gr.themes.Soft(),
    analytics_enabled=True
)

# Launch app
if __name__ == "__main__":
    interface.launch()