crisp_notation_cli 0.4.3

The crisp_notation command-line tool — inspect scores, convert MusicXML/MEI/kern/MIDI/MuseScore/GP/ABC, export LilyPond and braille, render SVG/PNG, and run OMR.

crisp_notation_cli

Command-line tool for the crisp_notation music notation libraries: inspect scores, convert between a dozen formats (MusicXML / .mxl / MEI / **kern / MIDI / MuseScore / the .gp tablature family / ABC, plus LilyPond and braille-music export), render to SVG or PNG (notation or tab), and recognize sheet-music images (OMR) — all pure Dart (OMR and PNG aside).

Usage

dart run crisp_notation_cli:crisp_notation <command> [arguments]

Command	Purpose
info <in>	Summarize a score (clef, meter, sizes, timeline length)
timeline <in> [--bpm N] [--no-expand]	Print the playback timeline (repeats/jumps unfolded unless --no-expand)
convert <in> <out>	Convert between any supported pair (formats inferred from the extensions, or --from/--to)
render <in> <out.svg> [options]	Render to SVG (notation, or --tab for tablature)
omr <image> <out> --model <smt.gguf> [--single]	Optical music recognition: a staff-notation image → score (.musicxml/.mxl/.krn) via the CrispEmbed Sheet Music Transformer

Input formats are inferred from file extensions — .xml/.musicxml, .mxl (zipped MusicXML), .mei (MEI), .krn/.kern (Humdrum), .mid/.midi, .abc (ABC notation), .mscx/.mscz (MuseScore), .gp (7/8) / .gpx (6) / .gp5 / .gp4 / .gp3 (and raw .gpif), and plain-text tab .tab/.crd/.txt — and can be overridden with --from (musicxml / mxl / mei / kern / midi / abc / mscx / mscz / gp / gpx / gp5 / gp4 / gp3 / gpif / asciitab). Output formats use --to or the output extension (.svg, .png, .mid, .musicxml, .mxl, .mei, .krn, .ly (LilyPond, export only), .brl (braille music, export only), .abc, .mscx, .mscz, .gp, .gpif). All formats funnel through one score model, so any pair round-trips transparently for the data they share. Plain-text tab is a lossy import; use --tuning to set the tuning (also for .gp), and --infer-rhythm to guess durations from the tab's horizontal spacing (otherwise all notes are eighths).

render options

Option	Meaning
--tab	Render as guitar/bass tablature
--tuning <std|dropD|bass>	Tab tuning (default std)
--staff-space <px>	Pixels per staff space (default 12)
--metadata <path>	SMuFL font metadata JSON (auto-located in the repo otherwise)
--no-embed-font	Do not embed the engraving font in the SVG

By default the engraving font is embedded via @font-face, so the SVG renders anywhere.

PNG: give the output a .png extension. Rasterizing needs a font rasterizer, so the tool delegates to the Flutter SDK (it runs flutter test tool/render_png.dart in the crisp_notation package, located automatically). The Flutter SDK must be on PATH; SVG needs only the Dart SDK.

omr — optical music recognition

Recognizes a staff-notation image (PNG/JPEG) into a score via CrispEmbed over FFI. The image is decoded in pure Dart; the engine is auto-detected from the model:

• Sheet Music Transformer → bekern tokens → a two-staff GrandStaff (or a single staff with --single);
• Polyphonic-TrOMR → semantic notation → a single polyphonic Score;
• Flova (handwritten / whiteboard) → LilyPond "simple notes" → an unmetered single-staff Score.

Output is .musicxml, .mxl, .krn, .svg, or .png — the last two render the notation (a grand staff for SMT, a single staff for TrOMR; --staff-space/--no-embed-font apply as in render). .png needs the Flutter SDK (it rasterizes via flutter test), like render.

Option	Meaning
--model <gguf|name>	OMR GGUF path, or a name that auto-downloads from Hugging Face to a cache — smt-grandstaff / tromr / flova (or set CRISP_NOTATION_OMR_MODEL)
--lib <path>	libcrispembed shared library (or set CRISPEMBED_LIB)
--single	Import only the first spine (single staff, not a grand staff)
--page	Full-page scan: split it into staff systems (horizontal-projection band detection) and recognize each, concatenating the results into one score
--threads <n>	Inference threads (default: auto)

Needs the native libcrispembed at runtime (built from CrispEmbed); the rest of the CLI is pure Dart.

Examples

dart run crisp_notation_cli:crisp_notation info song.musicxml
dart run crisp_notation_cli:crisp_notation convert song.musicxml song.mid
dart run crisp_notation_cli:crisp_notation convert song.gp song.musicxml    # .gp (7/8)
dart run crisp_notation_cli:crisp_notation convert song.mscz song.musicxml  # MuseScore
dart run crisp_notation_cli:crisp_notation convert song.musicxml song.mxl   # zipped MusicXML
dart run crisp_notation_cli:crisp_notation convert song.mei song.musicxml   # MEI
dart run crisp_notation_cli:crisp_notation convert song.krn song.musicxml   # Humdrum **kern
dart run crisp_notation_cli:crisp_notation convert song.musicxml song.ly    # LilyPond (export)
dart run crisp_notation_cli:crisp_notation convert song.musicxml song.brl   # braille music (export)
dart run crisp_notation_cli:crisp_notation render song.musicxml song.svg
dart run crisp_notation_cli:crisp_notation render song.musicxml song.png       # needs Flutter
dart run crisp_notation_cli:crisp_notation render riff.musicxml riff.svg --tab --tuning dropD
dart run crisp_notation_cli:crisp_notation render riff.tab riff.svg --tab      # import ASCII tab
dart run crisp_notation_cli:crisp_notation omr scan.png score.musicxml --model smt-grandstaff.gguf  # scan → score