If you built anything on Spotifys Audio Features API a mood visualizer, a workout playlist generator, a DJ key-matching tool you already know the bad news. Spotify shut it down. First the Audio Features and Audio Analysis endpoints in November 2024, then the Recommendations API in February 2026. Overnight, dozens of indie projects died. Spotify Pie, Discoverify, Moodify gone.
But the need didnt go away. People still want to understand their music. DJs still need harmonic matching. Playlist curators still want mood-based organization. The question is: where do you go now?
What Spotify actually removed
Spotifys Audio Features API provided 11 quantitative descriptors for every track in their catalog:
- Tempo BPM (beats per minute)
- Energy intensity and activity (0.0 to 1.0)
- Danceability how suitable for dancing based on tempo, rhythm, beat strength
- Valence musical positiveness (happy = 1.0, sad = 0.0)
- Acousticness confidence the track is acoustic
- Instrumentalness predicts whether a track contains no vocals
- Liveness detects the presence of an audience
- Speechiness detects spoken words in a track
- Loudness overall loudness in decibels
- Key the key the track is in (integer notation)
- Mode major (1) or minor (0)
These 11 features were the foundation of an entire ecosystem. Mood visualizers, playlist generators, recommendation engines, DJ tools all built on this data. When Spotify pulled the plug, they didnt just remove an API. They removed the lens through which an entire community saw music.
Why Spotify killed it
Spotifys reasoning was strategic, not technical. As they tightened Developer Mode access (now requiring Premium accounts and 250,000+ monthly active users), they consolidated their data advantage. The Audio Features API was a free gift to the market and Spotify decided the gift was over.
The move aligns with a broader industry trend: platforms pulling back from open APIs as they mature. Twitter did it. Reddit did it. Now Spotify. The era of free, open access to platform data is ending.
The alternatives in 2026
Heres whats left and whats actually good.
1. Self-hosted: Essentia and librosa
Essentia is an open-source C++/Python library for audio analysis. Its what Spotifys own researchers used before building their internal tools. It can extract everything the Audio Features API provided tempo, key, energy, danceability plus more (spectral analysis, rhythm patterns, tonal features). The catch: you need to host it yourself, process audio files, and handle the infrastructure.
librosa is the Python equivalent. Lighter, more flexible, but requires more manual work to extract meaningful features. Great for research, harder for production.
| Tool | Type | Cost | Best For | Difficulty |
|---|---|---|---|---|
| Essentia | Open-source library | Free (hosting costs) | Full audio analysis pipeline | High |
| librosa | Python library | Free | Research, prototyping | Medium-High |
| MeloData | REST API | Paid (usage-based) | Drop-in API replacement | Low |
| Moodwave | Product (Spotify integration) | $4.99 one-time | End users, playlist analysis | None |
2. Third-party APIs: MeloData
MeloData is the closest thing to a drop-in replacement. It uses the AcousticBrainz archive (3M+ tracks) plus on-demand Essentia analysis for tracks not in the database. REST API, straightforward integration. The downside: its paid, and coverage isnt as universal as Spotifys catalog was.
3. Purpose-built products: The Moodwave approach
This is where we built Moodwave. Instead of giving you raw data to process, Moodwave does the analysis for you. Connect your Spotify library, and it reads every tracks audio features tempo, energy, valence, danceability, acousticness and classifies them into 61 moods using music theory. Camelot Wheel harmonic matching. BPM flow optimization. Adaptive energy arcs.
The difference: you dont need to host infrastructure, write code, or understand signal processing. You just see your music differently.
The best API is the one you dont have to call. Give people the insight, not the raw data.
The market opportunity
The Music Mood Tagging AI market is projected to grow from $412.7M (2024) to $2.13B by 2033 at a 20.1% CAGR. The broader Music Visualizer Market is expected to hit $4.43B by 2035 at 27.8% CAGR.
Spotifys API shutdown didnt kill the market. It created a vacuum. The projects that died were built on free data. The products that survive will be the ones that built their own intelligence layer or partnered with someone who did.
What to build next
If youre a developer who lost your project, here are the paths forward:
- Self-host Essentia full control, full complexity. Best if you need custom analysis pipelines.
- Use MeloData fastest path to a working API. Best if youre rebuilding what you had.
- Build on Moodwave if your use case is mood visualization or playlist intelligence, weve already done the hard work.
- Go hybrid use Essentia for processing, store results in your own database, serve via your own API.
The audio analysis space is more open now than its been in years. Spotifys shutdown was a loss for the community but its also an invitation to build something better. Something that doesnt depend on a platforms goodwill.