Analysis
Audio features
Twelve measured dimensions per recording — rhythm, energy, mood, key and meter — from one call. The same numbers behind mood playlists, BPM-matched workout sets and content-based recommendation.
GET /audio_features?ids=…
What each feature means
Spotify derives these from the track's audio signal: a family of 0-to-1 confidence scores plus tempo in BPM, loudness in dB, and the harmonic and rhythmic frame. Together they describe how a song sounds independently of genre, language or release year.
Danceability
How suitable the track is for dancing, from tempo, rhythm stability, beat strength and regularity. Closer to 1 is more danceable.
Energy
Intensity and activity, from dynamic range, perceived loudness, timbre, onset rate and entropy. Death metal is high, a Bach prelude is low.
Valence
Musical positiveness. High sounds cheerful and euphoric; low sounds sad or angry. Paired with energy it forms the mood quadrant shown after analysis.
Acousticness
Confidence the recording is acoustic. Near 1 means primarily acoustic instrumentation with little electronic production.
Instrumentalness
Likelihood the track has no vocals. Above 0.5 is intended to read as instrumental; ooh and ahh sounds still count as instrumental.
Liveness
Probability an audience is present in the recording. Above 0.8 strongly suggests a live take rather than a studio one.
Speechiness
Presence of spoken words. Above 0.66 is mostly speech; 0.33 to 0.66 mixes music and speech, which is where rap sits; below 0.33 is music.
Tempo
Estimated speed in beats per minute, from the average interval between beats. Pop sits around 90 to 130; club and hardcore run well past 160.
Loudness
Average loudness across the whole track in decibels, typically between -60 and 0. Use it to normalize across a catalogue and avoid jumps on cross-fade.
Key and mode
Key is the estimated tonic in pitch-class notation (0 = C … 11 = B); mode is major (1) or minor (0). This page shows both by name — as integers they say nothing.
Time signature
Estimated meter, 3 to 7 beats per bar. Most popular music is 4/4; unusual meters surface prog, jazz and folk.
Duration
Track length in milliseconds, shown here as minutes and seconds. It is what every other per-second measure is normalized against.
Use cases
Mood playlists
Curate by the valence × energy quadrant — happy, calm, tense, sad.
Fitness apps
Match BPM to cadence; filter for high-energy tracks for intervals.
Sleep and focus
Low energy plus high acousticness is the ambient/study filter.
Taste matching
Score compatibility from two users' top-track feature profiles.
Music therapy
Build evidence-based sets for anxiety reduction and regulation.
Moderation
Flag speech-heavy tracks via speechiness plus the content rating.
Recommenders
Feed content-based filtering without touching listening history.
DJ tooling
Harmonic mixing by key and mode, BPM-matched transitions.
Deprecated by Spotify on 27 Nov 2024 — for whom
On 27 November 2024 Spotify removed Audio Features and Audio Analysis (with Related Artists, Recommendations, Featured and Category Playlists and 30-second preview URLs) from the Web API for new apps registered on or after that date and for existing apps still in development mode without a pending extension request. Apps that already held extended-quota access were not affected. The symptom is 403 on /v1/audio-features.
This endpoint answers the same question — twelve features per track — with one key. What it does not replace: the beat-by-beat segment grid of /v1/audio-analysis.
Before — Spotify Web API
GET https://api.spotify.com/v1/audio-features?ids=0VjIjW4GlUZAMYd2vXMi3b
Authorization: Bearer <OAuth access token>
→ 403 for apps registered after 27 Nov 2024After — this API
GET https://spotify-proxy.checkleaked.cc/audio_features?ids=0VjIjW4GlUZAMYd2vXMi3b
x-api-key: <your key>
→ 200, audio_features[] with the same Spotify track idEvery deprecated endpoint and its equivalent here, with dates and sources: the Spotify Web API alternative map. Source: Spotify for Developers, 27 Nov 2024.
FAQ
Why am I getting 403 from Spotify's /v1/audio-features?
Since 27 November 2024 Spotify no longer serves Audio Features to apps registered after that date, or to development-mode apps without a pending extension request. Apps that already had extended quota kept access. This endpoint answers the same request with one API key and no OAuth.
Is my app affected if it has extended quota?
No. Spotify's announcement states that applications with existing extended-mode Web API access relying on these endpoints remain unaffected. The change applies to new registrations and to development-mode apps.
Why ids and not id?
The endpoint is the batch form: /audio_features?ids=id1,id2 returns an audio_features array with one entry per id. Sending id returns 400 'Missing required query parameter: ids'.
Where do the scores come from?
Spotify's audio analysis pipeline extracts low-level descriptors — timbre, pitch, beats, segments — and aggregates them into the high-level features this endpoint returns.
Are scores stable across remasters?
A remaster gets its own track id and is re-analyzed, so values shift slightly. Danceability and key rarely move; loudness and energy do on a louder master.
Why does one track return no features?
Coverage is not universal — some regions and newer releases return an upstream error. The page says so rather than substituting a placeholder.
How do recommendations relate to features?
/recommendations?seed_tracks=ID returns tracks the model considers similar to the seed. Add target_energy, target_valence and friends to steer the result.
Build smarter music apps
Audio features, recommendations and popularity on one key. Free tier available.
Get your API key