Beat detection and song structure analysis API: the options compared (2026)

October 17, 2026 · 7 min read · every fact checked on the provider's own site on October 17, 2026

First, what you actually need back

Before you compare providers, decide which of these four you need. Half the "beat detection APIs" out there only do the first.

• Tempo: one number, "117.5 BPM". Enough for sorting a library or a BPM badge.
• Beats: the time of every beat. Enough for a beat-reactive visualiser or a cut every beat.
• Downbeats: the first beat of each bar. Needed when a cut, a light or a caption should land on the "one", which is where people feel a change.
• Sections: where the intro, verse, chorus and bridge start and end, ideally with a note on which sections repeat. Needed to find the chorus, skip the intro, or plan a video.

The table below marks each option on beats, downbeats and sections, then price and licence. The source link is the page the facts were read from on October 17, 2026. If a provider doesn't publish a price, the table says so instead of guessing.

The eight options at a glance

Each one in two sentences

Which one for which job

• A BPM badge or a sortable library: ReccoBeats (free) or librosa. You don't need beat times for that.
• A research project or a tool you run on your own machine: all-in-one for structure, librosa for beats. Both free, both yours to modify.
• Stems, chords and beats in one vendor: Music AI. Per-minute pricing, pick the modules you need.
• Tagging and searching a catalogue of thousands of tracks: Cyanite, or Sonoteller if lyrics metadata matters most.
• Cutting video, lights or captions on the beat for one song at a time: Songbrain. The grid comes with downbeats and repeat letters, and the same response tells you which 15 seconds to use and what's sung there.

What the Songbrain grid looks like

This is the real response for one of our own songs. Times are seconds from the start of the file. The full example is public at the URL above.

{
  "timeline": {
    "time_signature": "4/4",
    "beats":     [0.51, 1.02, 1.53, 2.04, ...],   // 350 beats
    "downbeats": [0.51, 2.55, 4.59, ...],         // 88 bars
    "sections": [
      { "start_sec": 0.0,   "end_sec": 16.37, "label": "A", "energy": 0.749 },
      { "start_sec": 16.37, "end_sec": 37.62, "label": "B", "energy": 0.934 },
      { "start_sec": 37.62, "end_sec": 57.91, "label": "B", "energy": 0.934 },
      ...                                          // 12 sections
    ]
  },
  "best_moments": [ { "rank": 1, "start_sec": 134.54, "peak_sec": 137.04, "end_sec": 149.54,
                      "kind": "outro peak", "score": 69, "reason": "...", "sung": "..." } ]
}

import requests, time
API, H = "https://api.songbrain.ai/v1", {"Authorization": "Bearer sb_live_YOUR_KEY"}

song = requests.post(f"{API}/songs", headers=H, files={"file": open("track.mp3", "rb")}).json()
while (doc := requests.get(f"{API}/songs/{song['id']}?include=timeline,best_moments", headers=H).json())["status"] == "processing":
    time.sleep(10)

beats, bars, sections = doc["timeline"]["beats"], doc["timeline"]["downbeats"], doc["timeline"]["sections"]
chorus = max(sections, key=lambda s: s["energy"])

Field reference in the API docs. Limits: 30 s to 10 min, 100 MB, MP3/WAV/FLAC/M4A/AAC/OGG/AIFF. The same tools are available to AI assistants through the Songbrain MCP server.