Create New Voice

Upload audio samples to analyze spectral features and train a new voice clone.

Upload Voice Samples

Voice Configuration

Recording Tips

  • 1
    Use a quiet, echo-free environment for best results.
  • 2
    Minimum 30 seconds of clean speech per sample.
  • 3
    Avoid music or background noise in your clips.
  • 4
    More varied speech samples produce better clones.

What we analyze

  • Mel spectrogram
  • MFCC coefficients
  • Spectral centroid
  • Fundamental frequency (F0)
  • Zero-crossing rate