RefBooth master analysis shows how your track will travel before release.

Thirteen readings in five groups, with fixes ranked.

Every reading outside its band is marked, and the report puts the first fix at the top. It also reads the tempo and key, and maps the track's sections.

Eight delivery targets, each labelled by its source.

Each target in the table below carries the source of its figure. Genre also moves the bands for dynamics, spectral tilt and bass, so an ambient record is held to different targets from a techno tool.

A platform's own published figure is used wherever one exists. Spotify normalises to -14 LUFS by default and asks for true peak below -1 dBTP, or below -2 dBTP for masters louder than -14 LUFS. Apple publishes no LUFS figure and recommends at least 1 dB of headroom.

Louder masters are turned down on streaming.

The major streaming services normalise playback. A master above the target plays quieter, so the extra loudness disappears and the compression used to reach it stays. AES TD1008, the Audio Engineering Society's streaming recommendation, recommends -16 LUFS for track-normalised music and a maximum true peak of -1 dBTP at the input of a lossy codec.

RefBooth shows the distance to each platform's target. A positive figure is how much that platform will turn the master down.

Why measure true peak instead of sample peak?

Because the waveform a listener hears can peak above the highest sample. A lossy encoder and a playback converter both rebuild the signal between samples, and ITU-R BS.1770 defines true peak for that reason, measured by oversampling at least four times for 48 kHz audio. Apple's 1 dB of headroom exists to avoid the clipping this causes.

RefBooth reports true peak and counts inter-sample overs. The target for overs is zero.

Uploads stay in Germany and train no model.

RefBooth's privacy policy (last updated 13 September 2026) commits to the following:

• Audio is measured on servers operated in Germany, inside the European Union, and nothing is replicated outside it.
• A file stays while its track is in your library, so a re-analysis needs no second upload. Deleting the track deletes the file. Deleting the account removes everything in one operation.
• Encrypted backups rotate out within six months and are never used to bring back something you deleted.
• No person listens to uploads, no other company receives them and nothing is used to train a model. The app has no analytics and no crash reporter.

RefBooth makes a different trade from a plugin meter.

Youlean Loudness Meter runs as a plugin inside one computer's session, and Loudness Penalty checks a file in your browser without uploading it. RefBooth uploads the file to its servers in the EU. In return, the whole report opens at one address and ranks what to fix first. For a file that must never leave your machine, use a local tool.

The standard report costs nothing.

It covers all thirteen readings against the platform you deliver to. Studio and Label plans are announced on refbooth.com, but the app has no checkout and asks for no card, so nothing can be charged yet. Once a plan goes on sale, its price, what it covers and how to cancel will be shown before the first charge.

RefBooth is one of the products we run ourselves. If the audio you need handled is phone calls, our voice and conversation systems transcribe them and file each one against the right customer.