← All work

VocaLattice

Local CLI · v1.0

A local-first audio-ML pipeline for vocal processing.

Stack

  • Python
  • Demucs (stem separation)
  • RVC
  • pedalboard
  • pyloudnorm
  • pytest

VocaLattice composes three open-source stages locally on CPU: Demucs isolates a vocal from a full mix, an RVC stage handles voice conversion, and a pedalboard chain (EQ, harmonics, reverb, loudness) polishes the result, with no API cost and no cloud.

A distinctive control: a SHA256-allowlisted model registry, so the loader refuses any model file whose hash isn't pre-verified. Built for my own music workflow.

Local CLI To verifyLocal-first audio-ML that stays on my machine, so there is no public link. I can walk through the code on request.

To verifyA link is held back until it's personally confirmed working and representative.