Index № notchlive March 2026 · Tiny Things Ship · 2026
NotchLive cover

NotchLive

Live captions for every Mac app. Private. On-device. In your notch.

NotchLive turns the area around the MacBook notch into a live captioning surface. Anything happening on the Mac (a meeting, a podcast, a lecture, a video, your own voice) can be captioned in real time, optionally translated, optionally saved.

How it works

Audio capture, speech recognition, and translation run as a single on-device pipeline. NotchLive uses ScreenCaptureKit to pull system audio from any Mac app, or the microphone for live conversation. Whisper runs locally through CoreML in five model sizes (Small to Pro), hot-swappable while captioning is live. Captions appear in roughly 1.5 seconds with partial updates. The notch rendering layer is shared with NotchPad so the foundation stays consistent across the suite. Macs without a notch get a top-center caption bar instead.

Translation runs entirely on-device via Apple Translation. 20 languages, no cloud, no API keys, no account. Whisper auto-detects across 90+ languages so the input language doesn’t need to be pre-picked.

What it is

Live captions for Zoom, Google Meet, Teams, YouTube, Spotify, Apple Podcasts, lectures, anything with speech. Built for hard-of-hearing users, non-native speakers in meetings, and anyone who wants spoken audio on a Mac to be readable in real time. No browser extension, no meeting bot, no account.

Pro adds voice notes (dictate with mic, keep the Raw transcript, optional AI cleanup, save locally or hand off to NotchPad), a saved transcripts library, session recording, and TXT/JSON export. $14.99, one-time, no subscription.

What I own

Product direction. Audio capture pipeline. Speech recognition stack and model-selection logic. Translation routing. Notch rendering and the top-center fallback. NotchPad hand-off. Native Mac app. Designed and shipped end to end as a Tiny Things product.

Built with

Swift on macOS 14+. ScreenCaptureKit and AVAudioEngine for system and microphone audio. Whisper via CoreML for on-device transcription, five model tiers selectable at runtime. Apple Translation for on-device translation across 20 languages. NSPanel rendered into the notch, with a top-center caption bar fallback for non-notch Macs. Sparkle for direct updates. Distributed via DMG (ScreenCaptureKit system audio capture is not compatible with Mac App Store sandboxing).

← Back to index