Runs locally No signup No watermark
Audio Tool
BPM & Key Finder

Two questions, two different answers under the hood. Tempo comes from a compact beat-tracking model (Beat This!, ISMIR 2024) that reads a mel spectrogram of the clip and marks every beat; the BPM you see is measured from the actual beat times, not from a lookup table, so half-time and syncopated material lands where a musician would count it. Key comes from no model at all: a chromagram folds the spectrum into the twelve pitch classes, and the profile is correlated against the classic Krumhansl-Schmuckler major and minor templates. When nothing tonal stands out, drum loops, speech, the tool says so instead of guessing. The report shows both results with a confidence for the key, and downloads as JSON with the first beat times included.

✓ Core Access ✓ No signup 🔒 Supported file contents are not uploaded for processing
1

Upload

2

Adjust & Preview

Pixlane
Preparing the tool on your device

Pixlane is loading on your device. Supported file contents are not uploaded for processing. Site analytics and the assistant are separate; tool-specific limits may apply.

BPM & Key Finder: Song Tempo and Key Online

Read a track's tempo and musical key in your browser. Beat-accurate BPM from a small on-device model, key from classical chroma analysis.

Two questions, two different answers under the hood. Tempo comes from a compact beat-tracking model (Beat This!, ISMIR 2024) that reads a mel spectrogram of the clip and marks every beat; the BPM you see is measured from the actual beat times, not from a lookup table, so half-time and syncopated material lands where a musician would count it. Key comes from no model at all: a chromagram folds the spectrum into the twelve pitch classes, and the profile is correlated against the classic Krumhansl-Schmuckler major and minor templates. When nothing tonal stands out, drum loops, speech, the tool says so instead of guessing. The report shows both results with a confidence for the key, and downloads as JSON with the first beat times included.

How to Use It

  1. Upload your audio file and wait for the local preview to load.
  2. Adjust the settings: Choose whether to report tempo and key together or one of them alone, then run the analysis; the audio is examined the same way either way.
  3. Download the result: Download the processed file directly to your device.

Controls and Options

Tempo + key covers most uses. Switch to BPM only when tagging a DJ library where key comes from elsewhere, or Key only when checking what to transpose a backing track to. Low key confidence means the material is weakly tonal. Trust the no-clear-key verdict over a forced label.

Formats and Output

Reads common audio formats your browser can decode, including MP3, WAV, M4A, OGG, and FLAC, and exports the analysis as a small JSON report with BPM, beat count, first beat times, key, and confidence.

Frequently Asked Questions

How accurate is the BPM reading?

The beat tracker is parity-locked against its research reference: on the validation clips the shipped engine lands within a few hundredths of a BPM of the reference implementation, and the count comes from real detected beats rather than autocorrelation guesses.

What does the key confidence mean?

It is the margin between the best-matching key profile and the runner-up. Above roughly 0.3 the key is clear; near the 0.15 floor the tonality is weak, and below it the tool reports no clear tonal center instead of naming a key.

Why does my drum loop get no key?

Percussion spreads energy across all twelve pitch classes, so no key profile stands out. That is the honest answer for atonal material, a key label there would be noise.

Does it handle half-time or double-time feels?

The model marks beats where the rhythmic emphasis actually falls, so a half-time groove reads at its felt tempo. If you disagree with the octave, doubling or halving the reported BPM is always the alternative reading.

How long a clip can I analyze?

Up to ten minutes of audio is analyzed in one pass; longer files are truncated to the first ten minutes and the report says so. Thirty seconds of steady material is usually enough for a stable reading.

Is my audio uploaded anywhere?

Supported media processing runs on your device and file contents are not uploaded for processing. Site analytics and the assistant are separate.

Back to audio tools