BRENNEN

Industries / Media, arts and recreation / Media and entertainment

Music recording

The music recording industry writes songs, records and mixes them, releases the recordings on discs, downloads and streams, and collects the money they earn for writers, performers, labels and publishers.

1877

Thomas Edison's tinfoil phonograph, built by his mechanic John Kruesi late in 1877, cut sound into foil on a turning cylinder and then played it back. The US National Park Service calls it "the first machine to both record sounds and play them back". Edison filed for a patent on 24 December 1877, and it was granted on 19 February 1878 as US patent 200,521, for a machine "to record in permanent characters the human voice and other sounds, from which characters such sounds may be reproduced and rendered audible again at a future time".

nps.govpatents.google.com

42

Milestones achieved across 6 domains.

17

To watch for: announced, not yet confirmed outside.

15

Open deliverables. Nobody has shipped them.

In one minute

Where it started

Edison's phonograph recorded sound and played it back in 1877. Writers had already organized to be paid: French composers and publishers set up SACEM in 1851 to collect money when their songs were performed in public. By 1898 companies were pressing and selling flat discs. Microphones replaced the recording horn in 1925, magnetic tape made recordings editable from 1948, and Ampex's Sel-Sync let a musician add one part at a time from 1955.

Where it stands

By IFPI's count, recorded music earned US$31.7 billion worldwide in 2025, up 6.4% and the eleventh year of growth in a row. Streaming brought in 69.6% of it, and 837 million people used paid streaming subscription accounts. Luminate counts an average of 106,000 new tracks delivered to streaming services each day in 2025, and Deezer says more than half of new uploads to its service were fully AI-generated by mid-2026. AI song generators are being brought under licence: Suno settled with Warner Music Group in November 2025 and launched models it says were trained on licensed music in September 2026, and in July 2026 a Munich court ruled that Suno's older models held copies of songs, in a judgment that is not yet final.

What is open

Much of the checking that pay, credit and charts depend on is still done by hand, or not done at all. Nobody has published a measured error rate for the free public links between recordings and songs, a table of the money Europe's collecting societies hold unpaid, a count of how often AI-made tracks carry the AI credit they should, or a test of whether playlist "bot checkers" predict Spotify's own corrections. Tape transfers and old Pro Tools sessions are checked by ear and by hand, and the claims for de-limiters, upmixers and lead-sheet models rest on their makers' own tests. Open data such as MusicBrainz and MUSDB18, open models such as SheetSage2 and the De-limiter, open audio-fingerprinting tools, and AI tools that read documents in bulk let a small team build and score these now.

Domains

AchievedWatch forOpen

Open in music recording

3 of 15 open deliverables

Open · Writing and producing

Demo to lead sheet, scored

A pipeline that takes a songwriter's rough phone demo, voice with guitar or piano, and writes a lead sheet with melody, chord symbols and lyrics lined up bar by bar, plus a public scorecard that compares each output with a lead sheet written by hand.

Built with what exists

SheetSage2 (open weights, CC BY-NC 4.0, so non-commercial use only), the model that makes this possible now; ElevenLabs Scribe v3 (see the Note "Fewer word errors") or OpenAI's open Whisper model for lyrics; Jam-ALT's alt-eval package for the lyric error score. For an optional public-domain set: US sound recordings from 1925, in the public domain since 1 January 2026 by Duke's Center for the Study of the Public Domain, scored against period sheet music from Johns Hopkins' Levy Sheet Music Collection.

Credit · unclaimed

Open · Recording

Test tape-transfer error detectors

An openly licensed test set of tape transfers in which every fault is known and logged by hand: wrong playback speed, wrong equalization (the treble and bass correction each tape standard needs), sections played in reverse, dropouts, and a tilted playback head. Add a scoring script, and publish scores for the MPAI CAE-ARP Audio Analyzer and at least one other method.

Built with what exists

Your own recordings dubbed to open-reel or cassette tape on a working deck; the MPAI CAE-ARP reference software (IEEE standard 3302-2022), whose machine-learning classifiers are the advance that makes automatic checking possible; MPAI's JSON "Irregularity File" format for the labels; SoX for scripted speed changes; the Padua tape-video dataset on Zenodo (DOI 10.5281/zenodo.14028922).

Credit · unclaimed

Open · Mixing and mastering

Score the unlimiters

A public scorecard for de-limiters. Score Ozone 12's Unlimiter, at default and maximum settings, the open De-limiter model, and no processing at all on every track of musdb-XL, a heavily limited copy of a standard test set, then run a small blind listening test.

Built with what exists

MUSDB18-HQ (150 songs with stems, for educational use only); the musdb-XL gain files (CC BY 4.0), which rebuild the limited copy from MUSDB18-HQ; the De-limiter code and weights, which its paper says are public; Ozone 12 Advanced; an SI-SDR scoring script and a BS.1770 loudness meter. The advance behind it is machine-learning de-limiting, published in 2023 and shipped in a product in 2025.

Credit · unclaimed

Where it sits