Score Scan Reader

heads: ledger checks on both binaries, narrow-font heads, fewer false hollow heads, front-end fallbacks (round 2) (2166f93)
site: changelog panel and engine label on /app/; ci: regression check against the last release on every push (03214a3)
symbols: tuplet detection with 840 divisions, robust key/clef reading on old prints, clef/key change detection (round 2) (bae11b0)

Earlier releases

demo: copista-2m ML detector demo on GitHub Pages (/ml-demo/) (f812157)
lyrics: split multi-note words, port strips line detector, read part labels and stanzas (round 2) (e1c2221)
parts: lyric-to-voice assignment, melisma extenders, printed part names, stanza credits (round 2) (20100ec)
site: release footer on every Pages page with version, release age (AEDT) and build commit (663c80b)
site: version bar at the top of every Pages page, including commits ahead of the release (e9c5c3a)
ui: read up to 40 pages page by page in a Web Worker with progress and Stop (round 2) (56859ef)

heads: ledger-line check, faint-staff detection, hollow-head recovery, wide-blob splitting (round 1) (177e0e8)
lyrics: wider word gaps, junk-row rejection, verse numbers and cross-system hyphens (round 1) (aca893f)
parts: barline-based piano detection, clef-based part names, per-measure rhythm repair, rest detection (round 1) (bcfb9e0)
publish the app, status page and docs on GitHub Pages (ee204e4)
score scan reader v1, eval harness, corpora, design docs, CI and release-please (ccbb2b6)
ship fix-forward round 1 (aggregate 0.589 -> 0.884) (d9c3916)
symbols: header parsing for clef/key/time, robust accidentals, treble-8vb clef (round 1) (7990370)

ci: use Playwright's bundled Chromium when the local cached build is absent (7d39275)

Drop a scanned PDF or image