The Canary Song Archive

The Canary Song Archive is a project that collects and publishes recordings of domesticated canary songs as an open dataset.

It brings together contributions from individuals to document how canary vocalizations vary over time, between birds, between breeds, environments, and everyday living conditions while creating a shared resource for both research and creative work.

Contact usListen to the files

Approach

Listening to variation as both evidence and expression.

Canary songs are learned and constantly changing. No two birds sound exactly the same, and even one bird's song can shift through routine, age, environment, and daily conditions.

The archive treats each contribution in two ways at once: as data, a measurable acoustic pattern that can be compared and revisited; and as expression, a living sound situated in a particular home, body, and moment.

Gathered across homes and contexts, these recordings begin to describe a wider field of repetition, difference, influence, and memory in domesticated birdsong.

Method

A simple distributed collection process.

The project is built through modest, practical contributions: a clear recording, a little contextual information, and permission to publish the file as part of the open archive.

  • Participants submit recordings made with a Raspberry Pi recorder, phone, or handheld recording device.
  • Each recording is accompanied by basic context such as location, time, and recording conditions.
  • Audio files are organized, preserved, and published for research and creative reuse.

Artistic use

Birdsong as material.

The recordings can become material for sound-based work, generative or algorithmic composition, listening experiments, and visual interpretations of acoustic pattern.

Data & privacy

Open data, transparent practices.

All recordings in this archive are contributed under the CC BY-NC 4.0 license and remain permanently available as open data. For full details on what is collected, how recordings are licensed, and your rights as a contributor, see the privacy policy.

Dataset snapshots

Monthly manifests for direct download.

The latest public manifest is 2026-05, with 991 recordings and 4.2 hours of audio.

The manifest is a JSON index of the public archive. Each entry describes one recording with the contributor username, bird and device identifiers, recording date, duration, location metadata, and links to the audio file, sidecar JSON, and spectrogram when available.

  • manifest.json is best for scripts, filtering, and direct file downloads.
  • metadata.csv is best for spreadsheets and quick comparison.
  • For automated downloads, use the urls.wav, urls.json, or urls.png fields.

Download the manifest, filter the entries you need, then fetch each file from its URL field. A simple script can loop through urls.wav to download the audio files and use the metadata fields to name or organize them.

Latest manifestLatest CSV