Browser-only match engine No audio fetched

Know it’s the right track before you press play.

Score every candidate result against title, artist, duration and album. Tune the evidence, set your threshold, and let the workbench say ACCEPT or NOT FOUND — with every sub-score exposed.

100% local Explainable 0–100 score No accounts · no upload

Step 1 · Evidence

Set the reference track

Reference title *

The track you actually want.

Artist *

Primary credited artist.

Duration

Optional. mm:ss or plain seconds.

Album

Optional. Helps separate versions.

Strictness

The acceptance gate. It never rounds up.

Candidate results *

One per line. Pipes, spaced dashes and pasted result lines are OK. Bracket noise is stripped before scoring.

Formats: Title | Artist | Album | mm:ss · Artist - Title | mm:ss 5 lines

Updates live. Nothing leaves this page.

Scoring your candidates…

The decision card will appear here.

Active weights will appear after the reference is complete.

Decision rule: only the top-ranked candidate can be accepted, and only if its total reaches the threshold. Everything below stays visible but gated.

Step 2 · Audit

Ranked candidates

5 scored · best first

Rows below the gate are muted but never hidden. Duration is missing-aware: if the reference has a duration and a candidate does not, that component scores 0 rather than being silently skipped.

Method

How the confidence score is built

Transparent by design: normalize, compare, weight, rank, then gate. No black box and no audio access.

01

Clean the text

Bracketed noise such as (Official Video), [Lyrics] and (Remix) is removed, along with feat. credits. Case, diacritics and punctuation do not decide the match.

02

Measure each signal

Title and artist use character and token similarity. Duration uses a tolerance curve: 0–1s is exact, 9s scores 0.73, and 31s or more scores 0. Album matching is a separate component.

03

Apply the gate

Scores are weighted on a 0–100 scale and sorted descending. Only the winner is eligible for ACCEPTED, and only when its total reaches your chosen threshold.

Default weights

when all fields exist

Title40%

Artist30%

Duration20%

Album10%

When your reference omits duration or album, their weight is redistributed across the remaining signals. A title-plus-artist reference is not punished for missing data you never supplied.

The three gates

LenientUseful for messy catalogs≥ 60

BalancedDefault for manual vetting≥ 75

StrictReject on any doubt≥ 85

The threshold is a hard gate, not a suggestion. The 2-decimal total is compared directly, so there is no free point from whole-number rounding.

FAQ

Does this search for or download music?

No. It never fetches audio, decrypts anything or contacts a streaming service. You paste candidate metadata; the scorer is a local decision utility.

What does the fuzzy score mean?

Each component is 0–1, where 1 is an exact normalized match. Title and artist combine character-bigram and token overlap, so typos and extra words degrade gradually instead of failing outright.

Why is a missing candidate duration scored 0?

If you supplied a reference duration, absence is a real lack of evidence. The weight redistributes only when the reference itself omits the field — not when a candidate fails to provide it.

Which candidate formats are accepted?

Can I tune the threshold?

Yes. Choose Lenient (60), Balanced (75) or Strict (85). The ranking and verdict update live, and the JSON export records the threshold used.

Is my data stored anywhere?

No. Every calculation runs in your browser. The copy button only puts the result JSON on your clipboard when you ask.