Drop a track.
Get its key, its tempo, and what mixes with it.

No, and you can check rather than take our word for it: open the network panel before you drop a track and watch. The file is read by the browser's own file reader, decoded by the browser's own audio engine, and measured by an analyser that is already sitting in the tab. There is no upload endpoint behind this page. The one request that does go out happens after the analysis and carries two values — a key and a tempo — because that is all the catalog needs to answer what mixes with it.

Measured, not asserted. Run over 1,000 tracks drawn at random from our catalog of 441,566 released records, this analyser agrees with the published tempo to within ±2 BPM on 87.6% of them, and gets the exact Camelot key on 64%. Another 9.3% land on the relative major or minor or the next key round the wheel — the two confusions every key detector makes, both of which you can still mix. Read those as agreement rather than accuracy: release metadata is entered by labels, and on tempo it is often a convention. 81.9% of the drum and bass we hold is published below 95 BPM, averaging 86.4, for music played at 170 to 180 — so when the detector and the label disagree by exactly an octave, which is 3.83% of the catalog, the grid is identical and only the counting differs.

Because a tempo detector locks onto the pulse it hears, and on fast material the pulse it hears is often half the written tempo. This is not a rare edge case — 10.5% of our own catalog is stored at the half-time reading, and 81.9% of the drum and bass in it. The tool offers you the other octave whenever there is a sensible one, and switching re-runs the catalog search at the tempo you chose. Neither number is wrong; they are two readings of the same grid.

Very likely both are defensible. Key detection is a judgement about which notes dominate, and a track with a strong bassline in one key and a pad in its relative minor genuinely sits between the two — that is why the relative major and minor confusion is the single most common disagreement between any two detectors. The confidence figure next to the key is worth reading: above 70% the reading is firm, below 50% the track is genuinely ambiguous and you should trust your ears. Where the two halves of a track disagree with each other, the tool says so rather than averaging it away.

MP3, WAV, AIFF, FLAC and M4A/AAC, plus anything else your browser can play. Chrome and Edge read every format a DJ collection contains. Safari and Firefox refuse some AIFF and FLAC variants — that is their decoder, not ours, and the page will tell you which file it could not open rather than failing silently. Files up to 500 MB, which is about twenty-five minutes of 24-bit WAV.

Where the audio stops. Every lossy encoder low-passes before it throws data away, and the corner it chooses is a fingerprint of the bitrate: around 16 kHz for 128 kbps, around 19 kHz for 256, around 20 kHz and up for 320 or lossless. The useful case is a file that claims to be lossless and is not — re-encoding an MP3 to FLAC keeps the MP3's cliff, and no amount of container says otherwise. If you have bought a "lossless" pack that stops dead at 16 kHz, you bought an upscale.

Select as many files as you like and they queue, one at a time, with every result staying on screen and a CSV export at the top once there is more than one. One at a time is deliberate rather than a limitation — the analyser holds a large working set in memory and running several at once is how a laptop tab runs out of it. For a whole collection the right tool is the library health check, which reads a Rekordbox export and measures the lot in under a second, also without uploading anything.

For a track already in the catalog, the key on its page came from Beatport's release metadata rather than from detection, so it is the label the artist and the distributor put on the record. This tool detects, which is what you need for a file whose metadata is missing or wrong — a promo, a bootleg, an edit you made yourself, or anything ripped without tags. The benchmark above is exactly the comparison between the two.

The analysis does, once the page has loaded — it is your machine doing the work. The catalog match is the part that needs a connection, since that is a question about our records rather than yours. Load the page at home, and the key and tempo will still come out in a venue with no signal.