Songs and genres

The library

109 annotated songs across 12 genres, built from real MIDI files. Fifteen of the MIDI files ship in the package; the other 94 are fetched from their source sites with one command (see the end of this page). America the Beautiful, arranged in this repository and dedicated to the public domain, is the fifteenth and the folk exemplar. Songs progress through three states:

1. Raw — MIDI only, no annotations
2. Annotated — musical language has been written by the AI
3. Ready — fully playable with bar-by-bar analysis and teaching notes

Genre exemplars

Each genre has one deeply annotated exemplar with historical context, bar-by-bar harmonic analysis, key moments, teaching goals, and performance tips (including vocal guidance). These serve as templates: the AI studies one, then annotates the rest.

Genre	Exemplar	Key	What it teaches
Blues	The Thrill Is Gone (B.B. King)	B minor	Minor blues form, call-and-response, playing behind the beat
Classical	Fur Elise (Beethoven)	A minor	Rondo form, touch differentiation, pedaling discipline
Film	Comptine d’un autre ete (Tiersen)	E minor	Arpeggiated textures, dynamic architecture without harmonic change
Folk	America the Beautiful (Ward)	F major	Tonic-dominant hymn harmony, patriotic phrasing, melody in F4–C5
Jazz	Autumn Leaves (Kosma)	G minor	ii-V-I progressions, guide tones, swing eighths, rootless voicings
Latin	The Girl from Ipanema (Jobim)	F major	Bossa nova rhythm, chromatic modulation, vocal restraint
New-Age	River Flows in You (Yiruma)	A major	I-V-vi-IV recognition, flowing arpeggios, rubato
Pop	Imagine (Lennon)	C major	Arpeggiated accompaniment, restraint, vocal sincerity
Ragtime	The Entertainer (Joplin)	C major	Oom-pah bass, syncopation, multi-strain form, tempo discipline
R&B	Superstition (Stevie Wonder)	Eb minor	16th-note funk, percussive keyboard, ghost notes
Rock	Your Song (Elton John)	Eb major	Piano ballad voice-leading, inversions, conversational singing
Soul	Lean on Me (Bill Withers)	C major	Diatonic melody, gospel accompaniment, call-and-response

Browsing the library

Use the Learn tools to explore:

• list_songs — filter by genre, difficulty, or keyword
• song_info — full musical analysis for any song
• registry_stats — library-wide totals (songs, genres, difficulties)
• annotation_progress — annotation status across all genres
• suggest_song — recommendation based on genre, difficulty, and play history

Annotation workflow

All 109 songs are now fully annotated — written by AI through this exact loop, gated by a quality rubric, and fact-checked against the actual MIDI. The same workflow promotes any newly imported song from raw to ready:

1. Use song_info or list_measures to examine the raw MIDI data
2. Study the genre exemplar for context and teaching patterns
3. Use annotate_song to write musical language — harmonic analysis, key moments, teaching goals
4. Use score_annotation to grade it; the song progresses from raw to ready once it clears the quality bar

Measure-level inspection

• list_measures — every measure’s notes, dynamics, and existing teaching notes
• teaching_note — deep dive into a single measure with fingering, dynamics, and musical context

Transposition

Transpose any song up or down by semitones using transpose_song. The tool shifts all notes and updates the key signature automatically. Useful for matching a singer’s range or practicing in different keys.

Section markers

Songs can have structural section markers (Intro, Verse, Chorus, Bridge, Coda) for navigation and teaching. Use list_sections to view them and add_section to annotate song structure.

Practice setup

Before playing, use practice_setup to get recommended speed, mode, voice settings, and the exact CLI command for a song. This factors in the song’s difficulty and the AI’s current skill level.

Use mute_hand to isolate left or right hand practice. Use preview_teaching_cues to see all teaching notes and key moments before playing.

What ships, and what you fetch

The annotations are ours and ship with every song. The MIDI files were downloaded from public MIDI sites when the library was built, and a per-file provenance audit found that only 14 of those downloads carry a licence that permits redistribution: Bernd Krueger’s piano-midi.de arrangements (CC-BY-SA-3.0-DE) and the Mutopia Project’s public-domain typesettings. Those 14 are in the npm package and in the repository. America the Beautiful is the fifteenth shipped MIDI: an arrangement made in this repository and dedicated to the public domain. The other 94 downloads are not. Each of their .json files carries a provenance block naming the source site, its terms and the file’s SHA-256, and

ai-jam-sessions library fetch --accept-source-terms

downloads each one from the site that published it, under that site’s terms, refusing any file whose hash no longer matches what the annotations were verified against. Until you fetch, those songs are listed as unfetched and the server plays the 15 that ship. Twelve files that turned out to be a different piece than their name were quarantined, which is why the downloaded library is 108 songs and not the 120 earlier versions claimed. America the Beautiful brings the library to 109. The full audit is in the repository at docs/findings/library-provenance-audit.md.