Voice Processor

import librosa
import soundfile as sf
import numpy as np
from dataclasses import dataclass

@dataclass
class ProcessingProfile:
    """Configuration for audio processing"""
    target_sr: int = 48000
    time_stretch: float = 1.0
    preserve_pitch: bool = True
    normalize: bool = True
    output_format: str = "wav"

class AudioPipeline:
    def __init__(self, profile: ProcessingProfile):
        self.profile = profile
    
    def process(self, input_path: str, output_path: str) -> dict:
        # Load audio
        y, sr = librosa.load(input_path, sr=None)
        
        # Time stretching with pitch preservation
        if self.profile.time_stretch != 1.0:
            if self.profile.preserve_pitch:
                y = librosa.effects.time_stretch(y, rate=self.profile.time_stretch)
            else:
                # Simple resampling approach
                y = librosa.resample(y, orig_sr=sr, 
                    target_sr=int(sr / self.profile.time_stretch))
        
        # Normalization
        if self.profile.normalize:
            y = librosa.util.normalize(y)
        
        # Export
        sf.write(output_path, y, self.profile.target_sr)
        
        return {
            'input_duration': len(y) / sr,
            'output_duration': len(y) / self.profile.target_sr,
            'peak_db': 20 * np.log10(np.max(np.abs(y)))
        }