Skip to main content
miniti can gently tap you on the shoulder during a live meeting if your delivery is drifting - a long monologue or a sudden filler-word spike. nudges are local notifications, computed on-device, with no llm and no network call.

the two nudges

nudgefires when
long monologueyou’ve been speaking continuously for ~60 seconds and have passed ~180 words without leaving a gap
filler spikemore than 8 fillers per minute over a rolling 60-second window (um, uh, like, …)
a 3-minute cooldown applies per nudge type, so you won’t get spammed. each one is a short, dismissable system notification.

which speaker is coached

  • macos - nudges are computed on your mic track (“you”). remote speakers from system audio are never nudged - the coaching is always about your delivery.
  • ios - capture is mic-only, so there’s no speaker separation at the audio level. nudges are computed against the single speaker heuristic miniti uses for ios, which in practice means phone-only capture can produce nudges that reflect whoever is nearest the mic. if you’re on a call via speakerphone with others, expect some noise here.

when they fire

nudges fire only when miniti is in the background. if you’re actively looking at the meeting view, you can already see the live metrics - a notification would just be noise. the moment you’re on the call (miniti minimized, a different app focused), the nudges kick in.

enable them

  • macos - settings (⌘,) → notifications → toggle nudge me when i’m monologuing and nudge me when i’m using too many fillers
  • ios - settings → notifications → same two toggles
both toggles are off by default. you’ll also need to grant miniti notification permission the first time you enable one.

privacy

  • fully local - detection runs on-device from the live transcript
  • no llm - it’s a simple word-count / rate calculation against a rolling window
  • no network - nothing leaves your device to produce a nudge
  • the notification itself is a local UNNotificationRequest scheduled by miniti

tuning

the thresholds (180 words, >8 fillers/min, 60-second window, 3-minute cooldown) are fixed in this release. if you want different values, email ian@miniti.app with your use case.