BPM Detector
Onset-energy autocorrelation estimates periodic pulse, not a certain musical beat. Half/double-tempo alternatives are common. Select a candidate and compare it by listening.
Related tools
- Tempo Tapper — Estimate a chosen pulse from recent tap intervals.
- Online Metronome — Practice a pulse with grouping, subdivisions and ramp.
- Song Key Finder — Rank major/minor tonal candidates from a local excerpt.
Estimate pulse candidates from a short local recording, then compare them by listening. The BPM Detector analyzes repeating changes in audio energy and lists likely periodic rates with relative strength scores. It helps you choose a working tempo, without claiming that one automatic number is the only musically valid beat.
Analyze a local excerpt
- Choose a readable audio file within 25 MB and 90 seconds.
- Select Analyze local audio and wait for the progress indicator.
- Review the candidate BPM buttons.
- Select a candidate and use Audition selected BPM.
- Compare that pulse with the original recording and save the candidate table as CSV if useful.
Begin with a section containing clear rhythmic attacks. A long ambient introduction, silence or a sustained pad may supply little useful pulse information. A short representative passage can be more informative than including every different section of a song.
What the algorithm is looking for
The method measures changes in short-time audio energy, then compares that pattern with delayed versions of itself. Strong repeating relationships create autocorrelation candidates. The search covers approximately 40–240 BPM and returns a small set of separated peaks.
This is a periodicity estimate rather than a full musical understanding of drums, phrasing or meter. A regular bass pattern, hi-hat subdivision or repeated chord attack can dominate the energy pattern even when you would count another pulse.
The strength score ranks candidates under this method. It is not a percentage probability that a BPM is correct, and a high value does not prove the song maintains that rate throughout its duration.
Resolve half-time and double-time candidates
A repeated rhythm can support rates such as 75 and 150 BPM. Autocorrelation can also emphasize longer multiples of a repeating period. Listen to how the phrase and backbeat fit each interpretation instead of selecting only the first or largest number.
A related candidate may feel more natural for rehearsal or editing even when it is not the strongest peak. The output is useful because it makes those alternatives visible. It does not determine an official tempo marking for the recording.
There is no automatic audio-speed change when you choose a candidate. The original file remains unchanged, and the audition is a separate short reference pulse. It is not a fully synchronized overlay or an exported click track for the entire song.
Check the beat unit before using the result
BPM counts a chosen beat. A song in compound meter can be described using eighth notes, quarter-note equivalents or dotted-quarter pulses depending on the task. The detector estimates periodic rates; it does not automatically assign the intended written beat unit.
For a simple duration check, 120 quarter-note BPM gives 500 ms per quarter. If you interpret that same numeric rate as another note value, the musical grid changes. Keep the intended unit with the BPM when entering it into a sequencer or effect.
The Tempo Tapper provides an independent manual pulse estimate. The Online Metronome provides longer practice playback once you have chosen the grouping and speed.
Why a result can be unclear or wrong
Music with weak attacks, changing tempo, rubato or several competing rhythms may not have one strong periodic answer. A no-candidate result means the method found insufficient qualifying repetition, not that the recording has no rhythm.
Stereo channels are averaged for this suite’s analysis. Opposing-phase content can weaken a useful pattern. The page does not isolate drums, remove vocals or provide a separate left/right selector.
A global excerpt estimate also does not describe every local change. Compare another short section when the performance varies. This tool does not generate a beat grid, bar-by-bar map or automatic tempo automation.
File handling and practical limits
Browser decoding determines which actual MP3, WAV, OGG, M4A or WebM encodings can be read. Mono/stereo and decoded-sample limits apply as well as compressed size and duration. Shorten a very high-rate file if it exceeds those limits.
The recording is decoded and analyzed locally rather than uploaded through this tool. Cancel interrupts analysis, and Reset clears the current state. Download CSV preserves the candidate BPM and autocorrelation values, not a processed audio file.
The result does not identify a song title or musical key. Use the Song Key Finder for tonal candidates, and the main Pitch Detector for one live fundamental. Those properties cannot be inferred from a pulse number alone.
Make the estimate reviewable
Keep the excerpt and selected candidate with your notes. Compare several measures and listen for gradual drift instead of checking just one transient. If two rates both fit, document which beat level you intend to use rather than disguising the ambiguity with more decimal places.
A useful BPM result supports a practical next step, such as choosing a rehearsal pulse or an editing grid. Its value depends on that application and a listening check, not only on the ranking of an algorithm score.
Frequently asked questions
Can it estimate BPM from an MP3?
Yes, when the local file decodes and fits the limits. Verify the returned candidates against the music.
Why do I see both a slow and a fast tempo?
Related pulse levels can fit the same repeating pattern. Half-time and double-time ambiguity is common.
Does the strength value mean confidence percentage?
No. It is an algorithm-specific autocorrelation measure used to compare candidates.
Will selecting a BPM retime the recording?
No. Use the separate Pitch Shifter’s tempo control when you want processed audio at another speed.