Music Studio · Reference

Spotify deprecated its Audio Features API. Here’s what to use instead.

On 27 November 2024, Spotify switched off the endpoints that quietly powered a decade of music tools. If your BPM finder, DJ app, fitness playlist generator or mood visualizer broke, this is why — and these are the honest options now.

What actually got shut off

For new applications, Spotify removed access to Get Audio Features (the famous 13-field vector: tempo, key, energy, danceability, valence, acousticness, instrumentalness, loudness, speechiness, liveness and the rest), Get Audio Analysis (beat, bar and section structure), Recommendations, Related Artists, 30-second preview URLs in multi-get responses, and access to Spotify-owned editorial and algorithmic playlists. Apps that already had extended access kept it; almost everyone else lost the data layer their product was built on. Spotify framed it as a security and anti-scraping measure.

For most teams, audio features were the product. Without them, there is no “sort my library by energy,” no “build me a 140-BPM running set,” no “match the visualizer to the track’s mood.”

The honest options, and what each is good for

ReccoBeats

The closest thing to a drop-in replacement: a free API that returns the same kind of feature metrics (danceability, energy, acousticness…) and can extract features from an uploaded audio file. Good for: teams that want the familiar Spotify-shaped vector back quickly. Watch for: reliability and coverage have been inconsistent in the wild — test against your real catalog before you commit.

Cyanite

A commercial music-analysis API focused on mood and emotion, offering everything Spotify exposed plus deeper tagging. Good for: funded products that need rich, reliable, emotion-aware tagging and can pay for it. Watch for: it’s a paid B2B service, not a free hobby API.

Essentia (self-hosted)

An open-source C++/Python library (with a JavaScript build, essentia.js) for audio analysis — tempo, key, beats, and high-level descriptors, run on your own machines. Good for: teams that have the actual audio files and want full control with no per-call cost or vendor risk. Watch for: you need the audio, plus the engineering to run and scale it. This is the most robust long-term answer if you own your files.

AcousticBrainz

A crowd-sourced acoustic database tied to MusicBrainz. Good for: looking up well-known recordings that were already submitted. Watch for: data collection stopped in 2022 — nothing released since then is covered, and there are no new submissions. Treat it as a frozen archive, not a live service.

GetSongBPM, SongBPM, Tunebat

Consumer-facing BPM and key finders. Good for: a quick human lookup of one song’s tempo or key. Watch for: limited or no official API, and independent testing has found their BPM/key values are not always accurate. Fine for a sanity check, risky as a data backbone.

A quick decision guide

  • You own the audio files → self-host Essentia. Most control, no vendor risk.
  • You want a free, Spotify-shaped API fast → try ReccoBeats, but validate coverage.
  • You’re a funded product needing reliable mood data → Cyanite.
  • You only need occasional human lookups → a BPM finder is fine.
  • You care what a song will feel like, not just its numbers → that’s the gap we built for.

Where Music I Want fits

We didn’t set out to clone Spotify’s feature vector. We were answering a different question — not what are this song’s numbers, but what will it do to a sensitive nervous system. So alongside tempo we track dynamic range, sudden changes, texture, predictability and vocal intensity, and roll them into a plain-language read of whether a song is gentle, moderate, or intense.

It’s a different lens, and an honest one: when we haven’t analyzed a song before, we estimate it from its known musical properties and say so. If you want the experiential layer rather than the raw metrics, the Music Studio is free and works on any song:

And for developers: our data is an open API

Everything in our database — BPM, dynamic range, sudden changes, texture, vocal style, misophonia triggers, sensory level — is queryable free, with no key and open CORS: GET /api/v1/song?title=…&artist=…. Full docs →

Last updated 1 June 2026. The deprecation took effect 27 November 2024. We keep this page current as the alternatives shift — if you maintain one of these tools and something here is out of date, the about page has a way to reach us.