Bach Doodle Dataset

dataset·user contributed symbolic music·active

User-entered melodies and model-generated four-voice harmonizations contributed through the 2019 Bach Doodle.

Verified facts

Official site	https://magenta.tensorflow.org/datasets/bach-doodle ↗
Geography	United States
size	sessions: 8500000 · music years: 6 · compositions: 21600000
scope	Miniature user melodies and Coconet harmonizations with interaction metadata.
genres	Bach-style chorale harmonization
license	CC BY 4.0
steward	Google LLC
creators	Cheng-Zhi Anna Huang; Curtis Hawthorne; Adam Roberts; Monica Dinculescu; James Wexler; Leon Hong; Jacob Howcroft
languages	not_applicable
modalities	MIDI NoteSequence; JSONL; TFRecord; country-level metadata; feedback metadata
geographies	Global
access method	Google-hosted JSONL or TFRecord archives and shards.
intended uses	music generation; music education; ethnomusicology
consent claims	Opt-in contribution described on the official page.
related papers	The Bach Doodle: Approachable music composition with machine learning at scale
prohibited uses	No specific prohibited uses stated beyond license compliance.
annotation method	Automatic session, key, timing, country-code, playback and feedback metadata.
collection method	Users created melodies in a web Doodle and could rate and contribute them to a public dataset.
provenance claims	Official page states users could choose to contribute compositions to the public dataset.
current availability	Official downloads active.
commercial use limits	No dataset-level noncommercial restriction stated; attribution required.
us market scope basis	Publicly available or materially used in US-facing music/audio-AI research.