Lilylet is a LilyPond-like music notation language designed for Markdown rendering and symbolic music representation in AIGC applications.
\key c \major \time 4/4 \clef "treble" c1 \bar "|."
[title "Jesu, meine Freude"]
[subtitle "BWV 610"]
[composer "J.S. Bach"]
\staff "1" \key c \minor \time 4/4 \clef "treble" \stemUp g'4 g f ef \\
\staff "1" \stemDown ef16[ d ef8]~ ef16[ f ef d] c8[ d]~ d[ c] \\
\staff "2" \clef "bass" c16[ b c8]~ c16[ b c g] a8[ g]~ g16[ g af ef] \\
\staff "3" \clef "bass" r8 c,16[ d] ef[ d ef8]~ ef16[ a, b g] c[ b c8] | % 1
\staff "1" \stemUp d2 c\fermata \\
\staff "1" \stemDown c8[ c4 b8] c8.[ \staff "2" \stemUp g16] \staff "1" c[ b c d] \\
\staff "2" f,16[ ef f d] g[ af g f] ef[ d ef8]~ ef16[ f ef d] \\
\staff "3" r16 g,[ af f] g[ f g8] c,2 | % 2
- Live Editor - Interactive editor with real-time music notation rendering
- Markdown Editor - Write documents with embedded Lilylet music snippets
- Tutorial - Comprehensive guide to Lilylet syntax
LilyPond uses a LaTeX-like text markup syntax with significant advantages:
- Beginner-friendly: Intuitive commands like
\clef,\key,\timerequire no knowledge of complex binary formats - Human-readable: Notes are represented directly as letters (c d e f g a b), durations as numbers (4 = quarter note)
- Relative pitch mode: Each note is calculated relative to the previous one—only octave shifts (
'or,) are needed when the interval exceeds a fourth, dramatically reducing octave markers. See LilyPond Relative Octave Entry
LilyPond is powerful but overly flexible—the same music can be written in multiple ways, which creates problems for AIGC scenarios:
| Issue | LilyPond | Lilylet |
|---|---|---|
| Verbose context | Requires \version, \header, \paper, \layout boilerplate |
Only core music content |
| Inconsistent formats | Relative pitch, absolute pitch, multiple chord notations | Unified format, reduced ambiguity |
| Complex nesting | \new Staff << \new Voice \relative c' { ... } >> |
\staff "1" ... |
- Shorter context description: Removes redundant information, allowing LLMs to process more music content within limited context windows
- Formatted layout: Fixed syntax structure facilitates model learning and generation
- Markdown-embeddable: Music snippets can be directly embedded in documents