Skip to main content
Lightning Jar - Web Studio Lightning Jar Wordmark

Study 28 · The chart-draft pipeline

The capstone chartless draft fails its gate and names the missing instrument: the comb's phase is a beat detector, not a downbeat detector

The chart-draft pipeline · the capstone

Study Overview

One Recipe, No Oracle

The grid used only inputs the feature has: the audio and the user-declared tempo and meter. The onset joint comb supplies the period (measured exact to 0.02 percent on all four songs since Study 24); the comb's own phase, registered at the lattice point nearest the first audible sound with the Study 25 physicality fence, supplies bar one; the audible span divided by the period supplies the bar count. One fable call per song on production-parity inputs (the shipped reader's hints and the transcription's bar notes over the derived grid), graded by root score against the chart, index-paired from bar one. eat-all-the-clocks graded against the corrected 71-bar chart.

The Gate Failed Where the Lattice Is Ambiguous

The comb cannot tell beat one from beat two or three: it maximizes onset energy at one lattice spacing, and on letter-lands and peaceful-dreams the backbeat carries more of it, so the anchor landed about 2.1 beats late and whole-bar index pairing then graded every bar against its neighbor's window. The drafts still held 81.3 and 89.5 percent, which measures how much root information survives a wrong phase, but the registered gate asked for within 5 points of the chart-given references and missed by 8 to 13. mop-heart, one beat late, drafted above its reference anyway; eat, downbeat-exact, drafted 80.9 percent against its 81.2 percent corrected-chart reference. The split is exactly Study 27's lesson inverted: v5's alignment placement is safe because the DTW line tethers the index and the half-beat fence bounds the slide, and the chartless construction has no tether, so the fence has nothing to hold on to.

What Passed, and What Ships Next

The length gate passed on all four songs, derived counts within one bar of the charts, so a recording plus a declared tempo recovers the song's bar length mechanically. The comb period leg is exact. The one missing instrument is downbeat disambiguation for the anchor, and the app's backbeat brief proposed its shape before this study measured the need: score the chosen grid against itself shifted by one and two beats and prefer the winner. That is the named next remedy for the auto-chart feature, a candidate Study 30, not yet registered. Spend: about a quarter dollar of a one dollar fence; predictions one wrong (the gate, on two songs), length and spend right.

Table 28.1: Per song: the chartless grid's anchor vs truth, the draft vs its reference, and the length inference

Ask Eljay

Ask about the studio's work, research, packages, or writing. A few starters:

Answers come from this site's own content and link their sources. For anything that matters, email hello@lightningjar.com.