AUDIO MASTER

Suno WAV → Commercially mastered track. Reference-based. Free. Better than Mixea.

What This Is

A local mastering pipeline that replaces $99/year services. Uses open source tools:

1. Matchering 2.0 — matches RMS, frequency response, peak amplitude, stereo width to a reference track
2. PyLoudnorm — verifies LUFS target is hit (genre-specific)
3. Pedalboard (Spotify) — optional final polish (saturation, stereo tweaks)

Why this beats Mixea: Mixea has 12 fixed presets and no album consistency. Matchering lets you use ANY reference track and batch-process whole albums with identical treatment. In blind tests, Matchering ranks 3rd out of 12 — behind only 2 professional human engineers.

Prerequisites

Python packages:

# Windows / macOS / standard venv:
pip install matchering pyloudnorm pedalboard soundfile numpy

# Debian/Ubuntu with an externally-managed Python (PEP 668), add the flag:
pip install matchering pyloudnorm pedalboard soundfile numpy --break-system-packages

(Or pip install -r requirements.txt from the skill directory.)

FFmpeg (REQUIRED — hard dependency, not optional): The LUFS correction + true-peak limiting stage (adjust_lufs in scripts/master.py) shells out to ffmpeg's two-pass loudnorm filter. This stage runs by default on every master, so the pipeline will fail without ffmpeg on PATH. Verify with ffmpeg -version. On Windows install via winget install Gyan.FFmpeg (or grab a build and add it to PATH); it is also already on the system from lyric-video-forge.

The Pipeline — 4 Steps

Raw Suno WAV  →  [1] Match to reference  →  [2] LUFS verify  →  [3] Polish (optional)  →  [4] Export

Step 1: Reference Matching (Matchering)

Takes your raw track + a genre-specific reference track. Matches:

• RMS (loudness)
• Frequency Response (tonal balance)
• Peak amplitude
• Stereo width

Output: 24-bit WAV that sonically mirrors the reference.

Step 2: LUFS Verification (PyLoudnorm)

CRITICAL: Tracks distributed via a distributor like DistroKid (Spotify/Apple/Tidal/YouTube Music) + lyric videos on YouTube all land on platforms that normalize to -14 LUFS.

This means mastering loud (-7 LUFS) is counterproductive — Spotify just turns it down and you lose dynamic range. The correct strategy is -10 LUFS with -1 dBTP ceiling — ~4 dB headroom above normalization floor for punch without sacrificing dynamics.

The math: Spotify normalizes everything to -14 LUFS. If you master at -10, Spotify plays it at -14 with a 4 dB gain reduction applied. If you master at -7, Spotify applies 7 dB reduction — same final volume, but your -7 master had to sacrifice 3 more dB of dynamic range to get there. Less dynamic range = worse perceived sound at the same loudness.

If outside target ±0.5 LU, script applies gain correction.

Step 3: Pedalboard Polish (Optional)

Apply final touches:

• Tape saturation (warmth)
• Stereo widening (for EDM/industrial)
• High-shelf boost (air, presence)
• Final brickwall limiter backup

Step 4: Export

Outputs:

• {name}_mastered_24bit.wav — hi-res for DistroKid
• {name}_mastered_16bit.wav — CD/streaming standard
• {name}_master_report.json — LUFS, true peak, dynamic range measurements
• {name}_streaming.wav — secondary version at -14 LUFS for Spotify/YouTube (optional)

Reference Track Setup

The skill needs genre-specific reference tracks. Place them in references/:

references/
├── reference_map.json          # Genre → filename mapping
├── industrial_metal.wav         # e.g. Rammstein "Mein Herz Brennt"
├── edm_industrial.wav           # e.g. Angerfist or Perturbator
├── doom_sludge.wav              # e.g. Electric Wizard "Dopethrone"
├── folk_metal_shanty.wav        # e.g. Wind Rose "Diggy Diggy Hole"
├── dark_cinematic.wav           # e.g. Two Steps from Hell
└── tribal_metal.wav             # e.g. Sepultura "Roots"

Rule: Reference tracks should be professionally mastered, commercially released, same key/tempo zone as your target (ideally), and match the energy you want.

CLI Usage

# Auto-detect genre from filename or flag
python scripts/master.py --input plateau.wav --genre doom

# Explicit reference
python scripts/master.py --input evek.wav --reference references/edm_industrial.wav

# Album mode — master multiple tracks with SAME reference for consistency
python scripts/master.py --album ./tracks/ --reference references/industrial_metal.wav --output ./mastered/

# Override LUFS target
python scripts/master.py --input track.wav --genre doom --lufs -8

# Include streaming version (-14 LUFS secondary output)
python scripts/master.py --input track.wav --genre doom --streaming

# Skip polish step (Matchering only)
python scripts/master.py --input track.wav --genre doom --no-polish

Pipeline Integration

This skill fits between lyric-forge (song writing) and lyric-video-forge (video):

┌─────────────┐    ┌────────┐    ┌──────────────────┐    ┌──────────────────┐
│ lyric-forge │───▶│ Suno   │───▶│ audio-master     │───▶│ lyric-video-forge│
│ (script)    │    │ (raw)  │    │ (mastered WAV)   │    │ (MP4 with audio) │
└─────────────┘    └────────┘    └──────────────────┘    └──────────────────┘
                                          │
                                          ▼
                                   DistroKid / Spotify

Scripts

Quality Gate

Before delivering a master, verify:

• LUFS within ±1 of target for the genre
• True peak ≤ -1 dBTP (no clipping)
• Dynamic range (LRA) ≥ 4 LU for doom/industrial, ≥ 6 LU for shanty
• No audible artifacts introduced by limiter
• A/B compared against original Suno output (should sound tonally closer to reference, but not completely different)