00 · Arrival
00:00.00
You have brought us a record.
Signal analysis precedes processing.
Velnor is a 32-stage deterministic mastering engine. Every stage has evidence. Every decision leaves a trace. Scroll, and watch it work.
01 · Listen
00:04.12
Stage 01 → 03 · Input conditioning
First, the signal arrives.
DC offset removed. Stray sub-audible energy swept. Gain staged to a known operating point. Only then do we begin to read what you have actually made.
- DC filter 20 Hz high-pass, 1st order
- Noise defense Inaudible-floor removal
- Gain stage Reference to −18 dBFS RMS
input−3.1 dBFS
rms−14.8 dB
dc−0.0021
sr48 000 Hz
ch2 / stereo
02 · Analyse
00:08.40
Stage 04 → 07 · Spatial, spectrum, adaptive scaling
We measure before we move.
The engine pulls twenty-seven metrics from your mix before a single gain stage fires. Not to impress you. To know, before any processing occurs, what it is actually looking at.
03 · Confidence
00:12.61
Stage 08 · compute_confidence()
Nine dimensions calibrate processing depth.
Every stage in the chain is gated by a weighted confidence score. On a clean mix, most of the engine quietly does nothing. When the evidence is strong, it acts at full depth. When it is ambiguous, it preserves.
≥ 0.75Full
adapted depth
≥ 0.50~75 %
depth
≥ 0.30~40 %
depth, cautious
< 0.30~15 %
depth, preserve
Genre certainty0.00
Archetype match0.00
Spectral certainty0.00
Dynamic certainty0.00
Transient clarity0.00
Tonal stability0.00
Stereo integrity0.00
Noise floor0.00
Low-end integrity0.00
Global confidence
0.00
computing…
04 · Low-band anchor
00:17.03
Stage 10 · 12 · ParametricEQ, MultibandCompressor
Your kick decides where low ends.
Fixed-frequency crossovers split at the same place on every record. Velnor finds the kick fundamental and sets the low-band boundary around it, so the sub and the body never argue about phase.
Generic tool: split at 120 Hz,
kick body cut in half
Velnor: anchor at kick
fundamental, body stays whole
05 · Masking suppression
00:21.40
Stage 17 · MaskingSuppressor · 24 Bark bands
Mud removed where it exists.
Mud removed where it exists.
Warmth kept where it matters.
Twenty-four perceptually-spaced critical bands. Suppression fires only where real masking is present. Your low-mid warmth is not carved out on the way past.
passive band · no action
< 0.8 dB suppression
> 1.5 dB suppression ·
evidence-led
06 · Codec round-trip
00:26.84
Measured · same source · 256k AAC encode
What survives the encoder.
A PCM master can reconstruct above its ceiling after AAC encoding, even when the limiter reading looked safe. The same source file was put through a 256k AAC round-trip and the decoded output was checked against a −0.2 dBTP hardware safety ceiling.
LANDR's Balanced/Medium preset targeted −11.3 LUFS; Velnor's streaming preset targets −14 LUFS, in line with platform normalisation. Both outputs were checked at the same −0.2 dBTP hardware ceiling.
LANDR · Balanced / Medium
AAC recon: +1.12 dBTP
Velnor · Pop preset
AAC recon: −0.20 dBTP
07 · How it's built in
00:31.18
Stage 23 · 24 · 29 · Clipper A · Clipper B · True Peak Limiter
A delivery-aware ceiling, with the check inside.
Velnor sets its limiter ceiling based on the target platform, then runs a simulated AAC encode internally. If the reconstructed peak crosses the safety gate, the chain corrects before the file is written.
True peak
delivery-aware limiter ceiling
−1.06 dBTP
Integrated loudness
requested delivery profile
−14.0 LUFS
Simulated AAC pass
checked against delivery gate
pass · 0 overs
08 · The ledger
00:38.02
Every stage, in order, as it fires.
Thirty-two stages. Every one accountable.
Nothing is hidden, nothing is inferred. Thirty fire on every master. Two stay dormant until you ask for them. This is the actual chain.
- 01DC FilterRemove sub-audible DC biasauto
- 02Noise DefenseSuppress inaudible-floor noiseauto
- 03Gain StageSet known operating pointauto
- 04Spatial AnalyserStereo correlation, M/S ratio, imageauto
- 05Spectrum AnalyserPer-band energy, tilt, centroidauto
- 06Spectrum LevellerGlobal broadband flatten to referenceauto
- 07Adaptive ScalingRule weights against genre profileauto
- 08Confidence System10-component score · gates every stageauto
- 09Spectral MatchDeviation converted to EQ bandsauto
- 10Parametric EQSurgical correction, kick-anchoredauto
- 11Parallel Harmonic VCAPhase-correct harmonic densityauto
- 12Multiband CompressorPer-band dynamics, kick-anchored splitsauto
- 13Transient ShaperAttack/sustain shaping per bandauto
- 14Stereo WidthTrue-stereo shuffler, mono-safeauto
- 15Mid / Side EQTonal shape separated from imageauto
- 16Dynamic CorrectionMacro-dynamic envelope repairauto
- 17Masking Suppressor24 Bark bands · evidence-led cutsauto
- 18Air & PresenceHigh-shelf lift, genre-awareauto
- 19Harmonic ExciterEven-harmonic warmth, restrainedauto
- ✦De-esserSibilance control · cross-stage awareopt-in
- 21Output TrimPre-limiter gain stagingauto
- 22LUFS AnalysisPlatform-aware loudness targetingauto
- 23Clipper · Stage ATransient-preserved bulk shapingauto
- 24Clipper · Stage BSoft-clip peak tamer, LRA-adaptiveauto
- 25Sub-Stability GuardLow-end instability vetoauto
- 26Low-Band TamerSub resonance controlauto
- 27User Bass LiftGuarded low-shelf · user overrideopt-in
- 28Transient Integrity GuardPre-limiter snap protectionauto
- 29True Peak LimiterCodec-safe ceiling, 4× oversampledauto
- 30Budget ReportChain ledger · every move, evidencedauto
- 31Sample-rate ConversionTarget-aware, linear phaseauto
- 32DitherTPDF dither at bit-depth targetauto
32 stages · 30 fire by default · 2 are yours to turn on.
The de-esser and the user bass lift wait for your call. The other thirty answer to the mix
itself.
You have read what it does. Now hear it.
Hear the reel.
Six records, each one run through the live engine. Gain-matched, ready for a proper A/B, and waiting directly below.