CoolFace
Apppublic

SaifuddinHanif/ai-music-engineer

sourceHugging Facemitupdated 6mo agoView on Hugging Face
0likes
App README

AI Music Engineer 🎚️

Autonomous end-to-end music post-production agent. Ingest → Analyze → Separate → Mix → Master → Deliver — fully automated or supervised.

What it does

UserInputOutput
Independent artistDemo WAV/MP3Clean mix + streaming master
Mix engineerMultitracks folderSession template + mix prints
Mastering engineerStereo mixLoudness-compliant masters per destination
Content teamStereo mixInstrumental, Acappella, TV mix, stems

Pipeline

9-stage autonomous pipeline with configurable oversight (L0–L3):

  1. 1.Ingest — validate formats, build project manifest
  2. 2.Analyze — LUFS, true peak, BPM, key, phase, spectrum, pops/clicks
  3. 3.Separate (optional) — 2/4/6/8-stem separation via Demucs
  4. 4.Session Prep — DAW template, routing, gain staging, bus structure
  5. 5.Mix — 5-pass mixing (balance → corrective → shaping → automation → print)
  6. 6.Pre-master QC — Gate C: headroom, mono, true peak
  7. 7.Master — loudness targets per destination + reference matching
  8. 8.Deliver — WAV 24-bit + MP3 320k + FLAC + running order + M3U playlist
  9. 9.Archive — snapshot, changelog, recall manifest

Usage

  1. 1.Open the app tab above
  2. 2.Upload your audio file(s) (WAV, AIFF, FLAC, MP3)
  3. 3.Set artist, title, destination (streaming / CD / club)
  4. 4.Click Run Pipeline
  5. 5.Download your masters from the Deliver stage output

Autonomy levels

  • —L0 — Assist only: produces checklists, no execution
  • —L1 — Supervised: pauses for approval at key creative gates
  • —L2 — Default: fully autonomous, interrupts only on QC failure
  • —L3 — Full autonomy: unattended batch processing

Technical

  • —Runtime: Node.js 20 (zero npm dependencies)
  • —Audio processing: ffmpeg
  • —Stem separation: Demucs (Meta, open source)
  • —Loudness targets: Streaming −14 LUFS, CD −9 LUFS, Club −6 LUFS