Drop a finished 2-track mix to begin. Everything runs locally in your browser — only measurements ever leave your machine.
Before you master this track
Two things only you know — they steer the AI more than anything it can measure. Both optional, and you can change them anytime under RELEASE TARGET.
Where's it going?
What's the song about?
What kind of session?
Master your track
Pick a path. Everything stays in your browser — only measurements ever leave.
Your source track
—
No connection detected. AI Master unavailable — use Match EQ or Manual.
AI MasterUltra · Analyst Visionrequires internet
From your source — choose how the AI sets its target:
Match EQPremiumworks offline
MasterVS
by LumanPOP
Make Master
SIM
OFF
off ▾
TARGET
—
LUFS · — ▾
LUFS
—
target —
PEAK
—
dBTP
DYN
—
crest dB
LRA
—
LU · RNG
STEREO
—
width
PHASE
—
corr
RELEASE TARGET
SONG INTENT — OPTIONAL
Only you know what the song is about. Pick a feeling (click again to clear) and/or add a one-line story — the AI weighs it. We never guess emotion from other tracks; this is yours to tell.
SIMULATION
DEVICE
CODEC
Off
CODEC
VOL
0:00.000
0 samp · — kHz
IN
OUT
LOOP REGION
0:00.000
of —
REGION—
⚠
CENTER · legacy .wrap renders here when cockpit is OFF; module stack lands here in Batches 7-8
How do you want to master?
Pick a starting point. You can change later.
Pick a Genre
Match your track's spectrum to a genre preset. AI Master will weight its decisions toward this genre too.
Pick a Reference
Drop up to three tracks you want yours to sound like. AI Master will shape the whole chain toward them, and Match EQ proposes Surgical EQ bands from the averaged target.
Pick what Match EQ should target. Original = no matching (chain still runs). Reference 1/2/3 = drop your reference tracks (1-3 — averaged when more than one). Genres = release-style presets (also set the loudness target).
Release target
A loudness goal to aim near — a dB or two off is normal, and platforms simply turn tracks up or down to match. Not a pass/fail limit. Custom sets the LUFS target without any physical-delivery constraints. The other targets carry mono/ceiling expectations the AI Engineer will respect when AI Master runs. Not affiliated with these platforms; targets are publicly documented.
Simulate on
Near Field Monitors (8″) — Flat 8-inch nearfield reference — honest low end, no coloring.
A rough approximation of a typical device — every real device differs. A sanity check, not a measurement, and never baked into the download.
live
Toolsadd to chain when your material needs them
Match ToOriginal
Match EQ
PREVIEW
Match EQ on Original — no matching active
Max bands6
Intensity100%
Match EQ is on Original — no matching active. Pick a Reference or Genre in the Match to row above. Match EQ will analyze the difference between your source and the target, then propose Surgical EQ bands to close the gap.
source
reference
delta (ref − src)
proposed bands
Surgical EQ
16-BAND PARAMETRIC
Click the graph to add a band · drag to shape · scroll over a node for Q
0ms latency
0bands active
Click the graph to add a band where you hear a problem. Drag nodes to shape · shift+drag for Q · alt+drag to snap · scroll over a node for Q · right-click to change type · double-click to bypass · ⌘-click to delete
1 kHz
+0.0 dB
No band selected · click a node on the graph to edit, or click empty space to add a band
Tilt EQ
SINGLE-CONTROL TONAL SLOPE
PIVOT ≈ 1 kHz · LOW-SHELF · HIGH-SHELF CASCADE
Tilt
0 dB
Single-control tonal slope, pivot ≈ 1 kHz. Negative = warmer (more low end). Positive = brighter (more air). Small moves go a long way.
Bass Mono
LOW-FREQ FOLD
VINYL · CLUB PA · MONO COMPATIBILITY
Cutoff
120 Hz
Many soft synths use intentional stereo bass — leave off if you want to preserve that.
Glue Compressor
BUSsingle-band · slow · cohesion
GR 0.0 dB
INPUT (dB)
OUTPUT (dB)
GR
Transient Shaper
M/S BUS
PUNCH · SUSTAIN
Side Compressor
M/S BUS
THRESHOLD · RATIO
Stereo Width
M/S MATRIX
WIDTH · M/S MATRIX
~46 ms latency
M/S Saturation
M/S2nd + 3rd · parallel mid+side
OUT −∞ dB
HPF 120 Hz
Multiband Compressor
FET3 bands · musical · big/air
GR 0.0 dB
Multiband Compressor
Modern4 bands · transparent · clean
GR 0.0 dB
Multiband Compressor
VCA4 bands · bus glue · mild color
GR 0.0 dB
Dynamic EQ
6 PROBLEM BANDS
FREQ · Q · THRESHOLD · MAX GR
Sensitivity50higher = lower threshold, more bands engage more often
Limiter
TRANSPARENT
TRUE-PEAK · LOOK-AHEAD · BRICK-WALL
Ceiling
−1.0 dBFS
Character
Release
Oversample
Release time
100 ms
Spectrum analysis
A · source
B · master
Simulation
reference
Vectorscope
vertical = mono · centred · phase-coherent. horizontal = side content · out-of-phase risk. amber trace = your master, post-limiter.
— corr · — balance
Signal
Loudness (LUFS)
vs streaming target
True peak
inter-sample · 4× OS
Dyn. range
crest factor
LRA
loudness range
Stereo
width
Phase
correlation
Original mix
unprocessed
LUFS
—
play to measure
Peak
—
dBFS
Dyn. range
—
crest factor
LRA
—
loudness range
Stereo
—
width
Phase
—
correlation
Your master ↓
download · simulation excluded
LUFS
—
play to measure
Peak
—
export ceiling −1 dBTP
Dyn. range
—
crest factor
LRA
—
loudness range
Stereo
—
width
Phase
—
correlation
Simulation
simulation · not exported
LUFS
—
play to measure
Peak
—
dBFS
Dyn. range
—
crest factor
LRA
—
loudness range
Stereo
—
width
Phase
—
correlation
Codec preview — hear your master after MP3 encoding
not encoded
Bitrate
B-side source
tap B above to switch from original ↔ master · tap WAV/MP3 here to compare clean vs encoded
128/192 kbps stress-tests hi-hats, cymbals, and reverb tails. 256/320 is closer to what listeners on premium tiers actually hear.
Export — bit depth, dither, format · download or save your settings
Bit depth
CD / streaming standard
Dither
applied at 16-bit · decorrelates quantization error into pure noise
ⓘ What's in the export?
Exporting Your master — Genre correction (7-band EQ) · Stereo width · Harmonic saturation · Glue compressor · Transient shaper · Side compression · True-peak limiter @ −1 dBTP.
Simulation (earbuds, car stereo, laptop, etc.) is for listening only and is NOT baked in to the download.
You have existing settings in the chain. How should AI Master handle them?
The AI has been told about your existing work and reasons around it either way — but only "Augment my work" actually keeps your bands in the final chain.
Encoding MP3…
rendering the master — a few seconds
✦ AI MASTER
AI Mastering
Analyzing
Typically 15–20s · building your master — playback opens when it lands