TRACKED VOCALS
The rulebook

The Standards

The Charter is what we refuse — commissions, paid placement, selling data. The Standards are what we require, and how we check.

One is ethics, the other is method. A contractor citing the registry to a client needs this page. A singer deciding whether to trust it needs both.

The Charter → · Noise floor (the method) →

The Legend

Decode any card on the registry from this block. Every badge, chip, and measurement has exactly one definition. A badge that cannot be defined in one sentence does not render on a card.

MEASURED

Computed from a file you supplied, by the same code that runs in the Vocal Reel Certifier and on the server.

If it was not measured, it is not printed as a number.

DECLARED

Stated by the person who manages the listing — the singer, or a parent or guardian.

Labelled as such. Never presented as verified, never inferred from another field.

VERIFIED

Checked by a human here against a source.

Unverified claims are stored and reviewable, and they are not printed.

Badges and Chips

The complete set of badges and chips that can render on a directory card, profile, or EPK.

Certified Reel
MEASURED

Every check the seal requires was measured and passed: inside the cap, true peak at or below 0 dBTP, no clipped runs, loudness measured.

SAG-AFTRA Cleared
DECLARED

The singer's stated union status.

Direct Buyout / Non-Union
DECLARED

The singer's stated union status — direct contract, buyout terms.

Scale · contractor negotiated
DECLARED

Union session: fees are negotiated by the contractor against SAG-AFTRA scale.

Dry Auditions
DECLARED

The singer provides unmixed, untuned dry vocal passes on request.

SINGLE PASS
DECLARED

How the Dry Lead was made — one continuous unedited take.

COMPED
DECLARED

How the Dry Lead was made — assembled from the singer's own takes without pitch, time, or mix processing.

Records remotely
DECLARED

The singer records and delivers audio remotely from their own space.

In-person sessions
DECLARED

The singer is available to track in-person at commercial studios.

Remote only
DECLARED

Shown when a singer declares remote delivery and not in-person sessions.

Parent-managed listing
DECLARED

A parent or guardian created and manages this listing; the booking contact is theirs.

Example listing

A demonstration listing. Its figures illustrate the scale and are not measurements.

Measurements & Capture Checks

Integrity checks and noise floor bands computed by the DSP analyzer from the raw audio file.

No clipping / True Peak

≤ 0.0 dBTP, 0 clipped runs

True peak measured per BS.1770-4 Annex 2 with 4x oversampling; flags three or more consecutive samples at 0 dBFS.

No hum / Mains Hum

50 Hz / 60 Hz + harmonics

Mains hum detected across the quietest windows. Audible on exposed cues at any floor level.

Steady / Gap Integrity

Window level variance

Intermittent bursts or fluctuating background noise across the quiet gaps in the take.

File floor

dBFS (10th percentile)

Measured from the quietest decile of the dry pass; includes printed preamp and tracking chain compression.

Room floor

dBFS (unprocessed)

Measured from an optional 5–10 second room-tone capture — the acoustic space itself, before any tracking chain.

Audible noise
> -48 dBFS

The floor is audible beside the voice, not behind it.

Library quiet
> -58 dBFS

A working floor. Printed compression places the file number lower than the room deserves.

Inaudible floor
> -70 dBFS

The floor sits under the voice. A producer will not hear it on a commercial cue.

Studio silent
< −70 dBFS

The silence between phrases is true silence.

The three tiers

Everything on a listing belongs to exactly one. If a fact is not in this table, it is not on a card.

MEASURED

A machine computed it

Computed from a file you supplied, by the same code that runs in the Vocal Reel Certifier and on the server.

If it was not measured, it is not printed as a number.

Reel lengthLUFSsample peaktrue peaknoise floorsample ratebit depth

DECLARED

A person stated it

Stated by the person who manages the listing — the singer, or a parent or guardian.

Labelled as such. Never presented as verified, never inferred from another field.

Voice typerange notestechniquescredit contributionstermsgeargenreslanguages

VERIFIED

A person here checked it

Checked by a human here against a source.

Unverified claims are stored and reviewable, and they are not printed.

Credits against label copy or cue sheets
FactTier
Reel lengthMEASURED
Dry Lead lengthMEASURED
Integrated loudness (LUFS)MEASURED
Sample peakMEASURED
True peak (dBTP)BS.1770-4 Annex 2. Full scale is a physical ceiling, not a preference.MEASURED
Clipped runs and edit discontinuitiesClipped runs gate the seal. Edit discontinuities on a dry lead flag against Single Pass claims.MEASURED
File floorQuietest decile of the dry pass, printed compression included. Public number. Never a seal gate.MEASURED
Room floorOptional unprocessed room-tone capture. Skipping it still publishes and still certifies.MEASURED
Mains-hum and intermittency flagsFrom the same 50 ms windows. They do not gate the seal.MEASURED
Sample rateMEASURED
Bit depthMEASURED
WaveformMEASURED
Certified Reel sealIssued when the polished file is measured, inside the cap, and clean at the peaks. Not for sale. A clipped reel still publishes.MEASURED
NameDECLARED
PhotoDECLARED
Location and time zoneDECLARED
Hub proximityDerived from the stated location, not GPS.DECLARED
Dry Lead style (Single pass / Comped)Self-reported take assembly. Contradicted by detected edits on a Single Pass claim.DECLARED
Voice type, also-covers, descriptorNever inferred from range notes.DECLARED
Range notes (low, belt, mix, falsetto, whistle)DECLARED
TimbreDECLARED
TexturesDECLARED
Techniques, vibrato on request, mix on requestDECLARED
Arrangement specialtiesDECLARED
SoundalikesDECLARED
GenresDECLARED
LanguagesDECLARED
Union statusDECLARED
Turnaround, session years, longest sessionDECLARED
AI guide policyDECLARED
Records remotely and in-person availabilitySelf-reported delivery modes. NULL means not stated.DECLARED
Travel and availabilityDECLARED
Work typesDECLARED
Gear (mic, preamp, interface, room)DECLARED
Mic / interface classRead from the named gear, not measured from the file.DECLARED
DAW and competencyDECLARED
Self-edit skillsDECLARED
Sight-readingDECLARED
Performance video linksDECLARED
Contribution notes on creditsSpecific parts or moments stated by the singer. Stamped DECLARED even when the credit itself is VERIFIED.DECLARED
WebsiteDECLARED
Booking contactShown because the listing exists to be contacted. On a parent-managed listing this is the guardian.DECLARED
Parent-managed listingAttested, not verified. Never a search filter.DECLARED
CreditsChecked against label copy or cue sheets. CLAIMED credits are not printed.VERIFIED

Voice types, in the order the site uses them: Soprano, Mezzo-Soprano, Alto (Contralto), Countertenor, Tenor, Baritone, Bass-Baritone, Bass, Basso Profondo. DECLARED — never inferred from range notes.

The reel standards

These numbers come from the same module the validator uses. If a cap moves, this ruler moves with it. The cap is a wall. The target is the teaching. Do not raise the target to match the cap.

Seconds, to scale

Vocal reel

target 45s · cap 70s

Dry Lead

target 20s · cap 45s

Extended Reel

cap 12:00 · no target, no seal

The long one — for the producer who already wants to keep listening.

Off this scale — 12:00 is more than ten times the polished wall. MP3 recommended. A ten-minute WAV is around 200MB, will not get a full measurement, and earns no seal. WAV is accepted only if it fits the existing download-size cap.

Vocal reel

Produced and mixed — the montage that sells the voice.

Back-to-back hooks, no intros, no instrumental runway. Lead with your strongest four bars — a contractor skimming twenty singers decides inside ten seconds.

Dry Lead

Your real voice (single pass or comped) — proof of what actually leaves your room.

Single pass or comped from your own takes. No Melodyne, no Auto-Tune — tuning or timing — no quantizing, no reverb, no mix processing. Your tracking chain is fine; a clean, broadcast-ready recording is exactly what this is for. Keep the tuning raw and the timing natural — how you push, pull, drive or drag a tempo is part of your voice, and this is the file that lets a contractor hear it.

Over the cap, the application is declined with a reason. Nothing publishes silently. Measured length is allowed 0.5s of encoder slack past the wall.

The Dry Lead

Your real voice. The line is pitch and time manipulation and mix processing — not your tracking chain.

Single pass

DECLARED

One continuous take, unedited. No cuts, no comping.

Comped

DECLARED

Assembled from the singer's own takes. Cuts and crossfades allowed; every phrase is still a real, uncorrected performance.

Fine

  • Single pass or comped from own takes
  • Your tracking chain — pre, comp, EQ
  • Real room, real voice
  • Any mic you own

Not this

  • Melodyne — pitch or time
  • Auto-Tune
  • Quantizing
  • Reverb or mix processing

The way you push, pull, drive or drag a tempo is part of your voice, and it is exactly what this file lets a contractor hear.

This is a declared term. The measurement cannot detect Melodyne or Auto-Tune, and we do not claim it can.

How the numbers are made

The noise floor is not a vibe and not a guess. It is five deterministic steps over the file you supplied — run the same steps yourself and you get the same number.

The floor measurement, end to end
  1. 1

    The pass

    The delivered dry pass, breaths and all.

  2. 2

    50 ms windows

    Cut into non-overlapping windows. RMS of each, unweighted.

  3. 3

    Sorted

    The same windows, ordered quietest to loudest.

  4. 4

    Quietest 10%

    Keep the quietest decile. Everything louder is discarded.

  5. 5

    One number

    Their mean, in dBFS. That is the floor.

One implementation

The same code runs in the browser Vocal Reel Certifier and on the server, so the two can never disagree about a file. Nothing uploads from the Certifier; listing intake measures the stored file on our side.

Loudness

EBU R128 / ITU-R BS.1770-4. K-weighting (high-shelf + RLB high-pass), 400 ms blocks, -70 LUFS absolute gate and −10 LU relative gate. Reported as integrated LUFS.

Sample peak

Highest absolute sample across channels. No inter-sample oversampling. Kept beside true peak — the difference is informative. The crest-factor floor withhold still uses this number.

True peak, clipped runs, edits

True peak: ITU-R BS.1770-4 Annex 2, 4× through the specified interpolation filter. 0 dBTP is a physical ceiling, not a preference. A clipped run is ≥3 consecutive samples at full scale; a one- or two-sample touch is not clipping. An edit discontinuity (dry take only) is a step discontinuity exceeding 12 dB with signal present on both sides (pre- and post-window median > −44 dBFS). Clipping gates the Certified Reel seal; edit findings on a dry take flag Single Pass declarations (remedy: declare Comped or supply an unedited take). Neither gates publication.

Sample rate and bit depth

Read from the file — WAV header for bit depth, the decoded buffer for rate. The Certifier treats these clocks as pro: 44.1 kHz, 48 kHz, 88.2 kHz, 96 kHz, 176.4 kHz, 192 kHz. The EPK checklist line for “48 kHz / 24-bit or better” is a house floor on that sheet, not a second seal.

Two floors

The file floor is the quietest 10% of 50 ms windows on the delivered dry pass — printed compression included. That is what a producer receives, and it is the number on the card. The room floor is the same measurement on an optional 5–10s unprocessed room-tone capture. Makeup gain raises the file floor dB-for-dB; only the room floor describes the space. A mixed reel is not a room. Printed compression raises the file floor dB-for-dB with its makeup gain. That is why we report the room and the file separately when a room-tone capture is present.

Two flags

Both reuse those windows. Mains hum: Goertzel at 50 Hz and 60 Hz and their harmonics; a peak 12 dB above the local spectral median flags. Intermittency: on a dry pass, windows within 20 dB of the floor (not a fixed fraction); 90th minus 10th percentile. On room tone, every window, 98th minus 10th. A spread of 12 dB flags. Both thresholds are starting values. The flags, not the number, catch a take that would be sent back. They do not gate the seal.

The bands

The capture rail runs from 0 dBTP to -80 dBFS. The left cap is the ceiling. The hatched stretch is signal. The four floor bands begin at -40 dBFS.

The span between true peak and the floor is the usable range — where the voice lives. A file at −1.2 dBTP over a −53.4 floor has 52.2 dB of usable range. That figure is a diagnostic for the singer and the Certifier. It is not printed on a public card.

The published rail — ceiling, signal, four bands
signalAudible noiseLibrary quietInaudible floorStudio silent
Audible noise
The floor is audible beside the voice, not behind it.
Library quiet
A working floor. Printed compression places the file number lower than the room deserves.
Inaudible floor
The floor sits under the voice. A producer will not hear it on a commercial cue.
Studio silent
The silence between phrases is true silence.

Why these boundaries

Recorded music has no industry noise-floor specification — the Recording Academy's Producers & Engineers Wing delivery recommendations, the closest thing to one, contain none. The −60 dB figures often quoted come from spoken-word delivery specs (audiobook, VO), measured on a few seconds of hand-picked silence in a finished file. We measure harder: the quietest tenth of a continuous pass, breaths and all — so our number reads conservatively next to theirs.

The boundaries follow the measured real world: professional studios typically sit at −70 to −80 dBFS, well-treated home rooms at −65 to −70, untreated rooms at −40 to −50 — and compression printed on the way in raises the file floor dB-for-dB with its makeup gain, by design. That is why we report the room and the file separately.

What this measurement cannot tell you. A quietest-decile floor is a steady-noise measurement. It is blind to anything intermittent — a passing truck, a door, a dog, radio bleed all hide in the loud windows — and it understates tonal hum, which is far more audible than broadband hiss at the same level. The two flags exist because of this, and they, not the number, are what catch a take that would be sent back. The floor does not gate publication and does not gate certification.

The full working definition — citations, what the measurement cannot do, how to reimplement it — is on /noise-floor.

The evidence, and the gap

Every link in the chain between a singer and a producer has a published specification, except one. That missing row is the whole reason this method exists.

LoudnessSpecified

A published algorithm, with a broadcast practice built on it.

ITU-R BS.1770-4 — Algorithms to measure audio programme loudness and true-peak audio level

Listening roomsSpecified

Background noise specified — but scoped to monitoring, not capture.

EBU Tech 3276 — Listening conditions for the assessment of sound programme material (§2.6)

MicrophonesSpecified

Self-noise published in dB-A for every studio mic sold.

Neumann — What is Self-Noise (or Equivalent Noise Level)?

Vocal captureNot specified

No noise-floor requirement. We could not find a published scale.

Recording Academy Producers & Engineers Wing — Recommendation for Delivery of Recorded Music Projects

The full table, with the actual figures and the quotations behind each row, is on /noise-floor.

What the seal requires

The Certified Reel seal issues itself. There is no application, no queue, and no price. It means the polished file was measured, inside the cap, and clean at the peaks. A listing publishes with or without it.

Certified Reel

  • Integrated loudness measured

    EBU R128, gated. Present, not merely attempted.

  • Length inside the polished cap

  • True peak at or below 0 dBTP

    Full scale is a physical ceiling, not a preference. BS.1770-4 Annex 2.

  • No clipped runs

    Three or more consecutive samples at full scale. One- and two-sample touches pass.

  • Measured by the same code the Certifier runs

    Browser and server cannot disagree about a file.

It is not sold, and it cannot be bought.

Replacing the file clears it until the new file is measured.

An Extended Reel never carries it — a long file is not fully measured.

A clipped reel still publishes. It does not carry the seal.

Publishing a listing

  • The three consents

  • Required fields

    Name, booking email, location, microphone, turnaround.

  • One reel measured inside its cap

  • Dry Lead style stated when dry lead supplied

    Single pass or Comped.

  • Guardian attestation

    Parent-managed listings only.

If a file is over the cap, the wrong format, or undecodable, the application is refused with a reason that points here. Nothing is published silently. A claim the machines cannot settle waits on a person at review — still not a silent publish.

Targets, for the teaching: polished 45s, Dry Lead 20s. The walls are 1:10 and 45s.

Changelog

Standards that move silently look arbitrary. These move in public, or not at all.

  1. Aug 2026

    Dry Lead split into two declared variants: Single pass (one continuous take, unedited) and Comped (assembled from the singer’s own takes, no tuning/time/mix processing). Edit-noise detector repaired to require signal on both sides of a candidate discontinuity (excluding phrase onsets), scoped to the dry take, and removed from the Certified Reel seal path. An edit discontinuity now blocks a Single Pass declaration (remedy: declare Comped or upload unedited take) rather than withholding a seal.

  2. Aug 2026

    Polished cap 60s → 70s. Sixty was chosen, not discovered; seventy gives real reels headroom. The 45s target is unchanged — the cap is a wall, the target is the teaching.

  3. Aug 2026

    Raw cap 30s → 45s, same reasoning. Dry Lead definition rewritten: the line is pitch/time manipulation and mix processing, not the tracking chain.

  4. Aug 2026

    Noise floor split into file floor and room floor. Bands renamed for audibility. Two flags (mains hum, intermittency) added. The public card no longer shows a ladder. The floor still does not gate the seal.

  5. Aug 2026

    Visual noise-floor meter restored on the card: marker on the −40…−80 scale, audibility band named. The number is still the reading; flags still earn “clean”; the floor still does not gate the seal.

  6. Aug 2026

    Certification now requires no clipping: true peak at or below 0 dBTP, no flat-topped runs. Previously the seal confirmed only that a file had been measured. Listings publish with or without it, as always.

  7. Aug 2026

    Capture strip on the card, profile and EPK: three measured checks (peaks, hum, steady) and a nameless floor meter. Named bands stay on the Certifier.

  8. Aug 2026

    Capture meter rewritten: a repair verdict above a 0 dBTP → −80 rail. Public cards fold the checks into the line. Usable range stays on the Certifier and /standards.

— Tracked Vocals · The figures on this page are imported from the validator. If they ever disagree, the validator is right and this page is a bug.