Some people do not hear themselves getting louder. In a restaurant, on a train, in an open office, the room fills up and the voice follows it. Sordina watches the gap between the two and taps your wrist, quietly, when yours has been well above the room for a while.

• 50 msbetween measurements
• 8 sminimum between two taps
• 4.99 $once, everything included

It measures a level, not a voice

The watch reads how loud it is, every fiftieth of a second, and keeps only a number. The sound itself is never copied, never stored and never recognised. Sordina cannot tell what you said, or that it was you.

It compares you to the room, not to a number

A library and a bar do not need the same threshold. Sordina spends five seconds learning the room, then keeps following it: a room that gets noisier raises the bar, while a loud voice does not, because voices leave gaps between words.

A peak is not a problem

A laugh should not buzz your wrist. It takes a second and a half of sustained loudness to earn a tap, and the counter falls back during the pauses.

It will not nag

One tap every eight seconds at the very most. A reminder that arrives too often stops being a reminder and becomes something you turn off.

You set how much room you want

Sensitivity from six to twenty-five decibels above the room, and a sustained window of one, one and a half or three seconds. What is too loud is not the same for everyone.

Nothing leaves the watch

No account, no server, no recording, no history. Listening only ever starts when you open the app. The privacy page says exactly what the microphone does and does not do.

Seven languages, nine variants: English, French, German, Spanish (Spain and Latin America), Italian, Japanese and Portuguese (Brazil and Portugal).

Sordina is at version 0.1. There is nothing to download yet — this page will carry the link when there is.