NotchLive Start Guide

How to set up NotchLive on Mac.

NotchLive dictates into supported Mac text fields, captions app audio and microphone speech, translates live captions, builds reviewable local meeting records, and can share enabled context with Codex or Claude Desktop through local MCP.

Updated for NotchLive 1.14.2

This guide follows NotchLive 1.14.2: the desktop workspace, live captions and translation, saved recordings, and Voice Notes you can continue later.

First-time setup

First launch opens a desktop setup window. Complete the permission and speech-model steps, then use the main workspace. The optional notch widget is a compact view of the same live work.

1. Install NotchLive. Open the DMG, drag NotchLive into Applications, and launch it from Applications or Spotlight.
2. Allow app audio. macOS calls this Screen Recording permission, even though NotchLive uses it to hear apps and media for captions.
3. Allow microphone access if you need it. This is optional for app audio, but needed for My Voice, Everything mode, and Voice Notes.
4. Choose your spoken language. Automatic recommends Whisper Small. For a specific language, setup recommends Apple Speech where supported and Whisper otherwise. Review the model and download status, then continue. The microphone initially uses the same language.
5. Decide on Meeting Intelligence. Pro users on supported Macs can turn on private meeting summaries during setup.

Apple speech requires macOS 26 or later and a compatible Mac and language. The Apple option uses Apple Dictation where Apple Speech is unavailable for a supported language. Automatic language detection and audio/video file imports use Whisper. Speech downloads are separate from Apple Translation and Apple Intelligence. Existing users keep Whisper after updating; change the engine in Settings → Speech & Audio.

Compare Apple Speech and Whisper with a familiar clip in one app.

If setup is not finished: NotchLive shows a Finish Setup action in the notch, menu bar, main window, and Settings. Use it to recover missing app audio permission or a missing speech model.

Current new-install defaults: NotchLive starts in Everything mode with Timeline, a 40-point floating caption size, the top-center app switcher off, and Launch at Login off. Translation follows the Mac's preferred supported language, microphone language initially follows the language chosen during setup, and skipping microphone access leaves captions in Apps & Media mode. Existing users keep their preferences.

Home and workspace search

Start from Home to begin live captions or import an existing recording. The sidebar opens Live captions, Sessions, Meetings, Notes, Dictations and Settings. Reading existing saved content, recent lists and workspace search are free; creating new recordings and Voice Notes requires Pro.

Click Search or press ⌘K while the NotchLive window is active. Search across saved sessions, meetings, notes and dictations, or select a scope. Use the arrow keys and Return to open a result; Escape dismisses search.

Workspace search finds saved work. ⌘F in the caption window searches the current Timeline instead. To separate the next conversation, choose Start fresh… in Timeline and confirm: current captions and translations clear, an active recording stops and saves, and the current meeting summary saves. Existing sessions stay unchanged; captions retain their running or paused state.

Insights

Open Insights from the sidebar, Home, or the command palette (⌘K). It is free and summarizes existing saved sessions, recurring meeting topics, extracted decisions and action items, and recordings added to earlier notes. Choose a month or all saved history, then open a source to revisit the original.

Caption activity adds daily active time and finalized-caption counts from version 1.13.3 onward. These local counters save no transcript text or audio. Active time includes silence during capture and excludes pauses; it is not your speaking time. Opening Insights runs no new AI analysis.

First captions

Open a Zoom call, Google Meet, Microsoft Teams meeting, podcast, lecture, video, or any other Mac audio source. Then start captions. For the broader product view, see Mac live captions for calls, videos, and Mac audio.

• Press ⌥⌘C to start or stop captions.
• Use the NotchLive menu bar item if you prefer clicking.
• Hover over the notch to reveal controls for mode, recording, display, audio source, Ghost Mode, and Settings.
• Press ⌥⌘K to copy the latest caption.

Floating caption window

Open Caption Display when you want larger captions, a different position on screen, or space to read back through the conversation. It uses the same caption stream as the notch.

• Choose Open Caption Display in the menu bar, use the notch’s display control, or press the default shortcut ⌥⌘F.
• Move it by the title bar and resize it from an edge or corner.
• Use A− and A+ to change floating caption size.
• Use the lock-shaped Keep Window on Top control to keep captions above ordinary working app windows. This does not cover every macOS system surface.

Close is not Stop: closing Caption Display hides the window, but active captions continue. Use the caption start/stop control or ⌥⌘C to stop capture.

Read the floating caption window guide →

Focus mode

Select Focus at the top of Caption Display for a compact view that emphasizes the latest speech. It is a NotchLive reading view, not macOS Focus or a notification-blocking mode.

Choose your audio source separately: Apps & Media for Mac playback, My Voice for your microphone, or Everything for both. Switching to Focus does not change that choice or restart capture.

When you need earlier context, select Timeline. Pressing ⌘F while the window is active also switches to Timeline and opens search. Focus and Timeline browsing/search are included for free.

Read the Focus mode guide →

Live Timeline

Select Timeline in Caption Display to read back through the current caption run while new captions continue arriving.

• Scroll upward to recover a detail, then use Jump to Live to return to the newest captions.
• Use ⌘F to search finalized captions in the current run. It cannot find audio from before capture started.
• With Pro, use Copy for the full current transcript or Export for a TXT or Markdown snapshot.

Live context is not a saved meeting: an unrecorded Timeline stays in memory for the current run and is not restored after quitting. Opening Timeline does not start recording. Use Record with Pro when you intentionally want a retained session.

Read the Live Timeline guide →

Audio sources

Settings uses clear audio-source names, while compact notch controls may use shorter labels. You can switch sources from either place.

Apps & Media

Captions audio from Mac apps: meeting apps, browser video, podcasts, lectures, music, and other system audio. This is the best default for most people.

My Voice

Captions microphone speech only. Pick a Microphone Language if you mostly speak one language and want better microphone accuracy.

Everything

Captions app audio and your microphone together. Use it when you want your own speech and other speakers captured in the same workflow.

Open Settings → Speech & Audio. Set Apps & Media Language and Microphone Language independently; Auto is useful for mixed speech, while a known language can help recognition. These describe the speech you hear, not the translation you want to read.

For microphone captions, choose Kōkako or Legacy. Stop captions before changing the engine, then resume. Kōkako offers Live Caption Sensitivity: Quiet Speech, Balanced or Noisy Room. Voice Notes use Legacy in this release; see the Kōkako guide for the playback-reference requirements.

Speech models: open Settings → Speech & Audio, download the model you need, then select it. Finish any active recording or Voice Note before switching. If the app reports missing or changed files, use Repair Model in this category; repair preserves saved work and settings.

Translation Pro

Translate live captions on device with Apple Translation on compatible macOS and language pairs. For the full translation workflow, see the Mac live translation app.

• Open Settings → Translation and set Mode → Translate. The notch mode control or ⌥⌘T also changes mode.
• Set Translate To to the language you want to read. Source recognition languages are set separately under Speech & Audio.
• Under Live translation, choose Faster updates for provisional text that can change, or More context to wait for a completed caption in the full Timeline window. The compact view and notch use Faster updates.
• Enable Show original with translation to keep the recognized speech beside its translation.
• In Translation Languages, choose Set Up… for the required pair. In the macOS sheet, click every requested Download, wait, then click Done. Use Retry after a failed setup; Set Up does not itself finish the downloads.
• Built-in translation requires Pro, macOS 15 or later, and a supported language pair. Recognition happens first; timing and quality depend on the audio, model, language pair and Mac.

Read the translation setup guide →

Dictation

Dictation turns microphone speech into text in supported standard, non-secure Mac text fields. Recognition, Personal Dictionary rules, recent-history recovery, and insertion coordination stay on your Mac.

1. Dictation is free. Enable it in Settings and allow microphone and macOS Accessibility access to insert text into the focused field.
2. Hold ⌥Space and release to insert. In Settings → Dictation, choose Fn / Globe or record a custom shortcut. Toggle mode starts with one press and finishes with the next. For Fn / Globe, set “Press Globe key to” to “Do Nothing” in System Settings → Keyboard.
3. Translated Dictation requires Pro and macOS 15 or later. Use ⇧⌥Space to translate from your microphone language into the target language selected in Settings.
4. Open Dictations for recovery. Search recent results, inspect original or translated text, and copy an entry if the original field could not accept insertion.
5. Manage the Personal Dictionary. Add vocabulary for recognition context or deterministic spoken-to-written replacements for names, acronyms, and specialist terms.

In Settings → Dictation, set Translate dictation to separately from the live-caption target, use Set Up for its language pair, and choose the Dictation bar position. Shortcut Behavior selects hold-to-speak or press-once operation. Personal Dictionary and recent Dictations are available from this category too.

Personal Dictionary example: Settings → Dictation → Personal Dictionary → Manage… opens the editor. Add “NotchLive” under Preferred spellings as a recognition hint. For a fixed replacement, enter “project aurora” in When I say and “Project Aurora” in Write, then Add. Replacements run locally after recognition and before translation. Bar position offers Automatic, top/bottom center and screen corners.

Meeting Intelligence Pro

Meeting Intelligence can analyze live captions without a session recording. It requires Pro, macOS 26 or later, and an available, enabled Apple Intelligence model. Recording is a separate choice when you want to retain a session transcript and optional audio.

Meeting Intelligence

On supported Macs with Apple Intelligence enabled, NotchLive can turn captions into current focus, key points, open questions, action items, key moments, and saved meeting summaries.

Ask-to-record prompts

In Settings → Sessions & Meetings → Meeting Detection, choose the ask-to-record option if you want prompts for likely calls. It never starts recording for you. The meeting visibility choice controls whether the notch is shown or kept hidden.

Export and share

Saved Meetings, Sessions, and Notes support native sharing, copy, export, and delete actions from the NotchLive window.

Open Settings → Sessions & Meetings for Auto-run Meeting Intelligence. Also auto-run in Translate mode is a separate option and may slow translation. With it off, automatic analysis pauses in Translate and catches up in Caption mode; you can still start analysis manually. Saved summaries appear in Meetings and link to a session when one was recorded.

Manual analysis: open Meeting Intelligence from the live toolbar. When it is paused in Translate mode, the panel offers Run anyway. This is separate from recording a session and from enabling automatic analysis.

Voice Notes Pro

Voice Notes records your microphone into a local note with audio. Stop saves it; later recordings can continue the same note, keeping dated clips and their text together.

• Choose Notes from the NotchLive mode controls.
• Tap Record and speak naturally into your microphone.
• Press Stop to save the note and its microphone recording. Recordings use the Compact or Lossless quality selected in Settings.
• Use Clean, Translate and Raw in the active Notes view. The saved-note detail has recording, playback and export controls; it does not have that view switch.
• Open a saved note and choose Continue recording to add more later. Each recording keeps its own date; use its menu to rename the label.
• Play individual recordings or use Play all. Export clips separately, or choose Export combined audio for one M4A file. Multiple recordings are encoded as Compact; a single existing M4A is copied unchanged and can remain Lossless.
• Copy, export, share, or send the note text to NotchPad.

Read the recording and continuation tutorial →

Saved work, audio and backups

• Record a session: start live captions, choose Record with Pro, and stop recording to retain the transcript in Sessions. Opening Timeline alone does not save a session.
• Retain session audio: in Settings → Sessions & Meetings, enable Save audio with sessions and choose Compact or Lossless. This is off by default and applies to session recordings. Voice Notes save microphone audio when you record a note independently of this switch.
• Import an existing file: choose Import from Home. The source file stays unchanged; video imports retain audio for playback. Follow the audio and video import guide.
• Review speakers: open a saved session's participant inspector. Hover a participant for Rename or Merge participant; click a transcript speaker label to reassign a passage, leave it unknown, or use Move selected words… for a smaller correction. Automatic labels are evidence to review, not identity recognition.
• Translate a saved transcript: open its translation control, set From and To, then choose Translate transcript or Pause captions & translate. Original remains intact; each language tab has its own search, copy, share and export. Saved audio is not required.
• Export and back up: Settings → Data & Exports → Default Folder → Choose… selects a folder before Auto-export sessions and notes can be enabled. It writes TXT, Markdown and JSON in item subfolders, with combined M4A for sessions containing audio. Export All Data… creates a library bundle; Import Data… restores or transfers one without overwriting conflicting data.

Ghost Mode

Ghost Mode hides NotchLive from screen sharing and screen recordings while captions keep running locally on your Mac.

Enable Ghost Mode in Settings → Privacy or use the ghost control in the live toolbar. Check your sharing app's preview before a presentation.

Settings

Open Settings from the workspace sidebar. Search Settings to jump to a control, or use its categories. The reference below groups the smaller options by their actual location.

Useful shortcuts

About Sign Cues: Sign Cues are ASL-inspired visual markers for selected caption moments. They are experimental and are not sign language translation or interpretation.

Pause after inactivity

NotchLive 1.9.2 can pause capture after 15, 30, or 60 minutes without audio activity, or stay active indefinitely with the default Never setting. Recordings and Voice Notes are never paused by this setting. Use Wake or Retry if the caption state reports that capture is sleeping or unavailable.

Small settings worth knowing

• Appearance: Window Theme offers Follow System, Light or Dark; Dark Shade and Light Shade choose the background style. Enable Transparent Background to reveal its transparency slider; text stays visible. Set caption sizes, wave color and Caption Alignment: Centered or Natural, which follows the text’s direction.
• Appearance → Show microphone captions below widget: hides the secondary panel; it does not stop microphone recognition.
• Appearance → Hide after silence: changes caption visibility. General → Pause after inactivity changes capture; its Never default and 15/30/60-minute options do not pause recordings or Voice Notes.
• General: configure Launch at Login and top-center hover. The app switcher appears when NotchPad is installed; direct NotchLive hover is available when the switcher is off or unavailable.
• Appearance → Sign Cues: experimental visual cues with name, size, speed and contrast controls. They are not sign language translation.
• Shortcuts: record your preferred keys. Dictation has its own hold or toggle behavior and bar placement under Dictation.
• License: paste the purchased license key and choose Activate; use Deactivate before moving an activation when needed.
• Privacy / About: control anonymous diagnostics under Privacy; check for updates under About. Diagnostics do not include your audio, captions, transcripts or notes.

AI Connections and MCP Pro

Codex, Claude Desktop and Claude Code have separate setup rows in Settings → AI Connections. Claude Code setup updates only its user-level MCP connection.

NotchLive 1.9.1 runs MCP through a dedicated headless helper with no Dock icon or app window. It can configure Codex, Claude Desktop or Claude Code directly, and it provides a standard stdio configuration for Hermes, OpenClaw, compatible Obsidian plugins, and other MCP-capable hosts.

Verify the connection

Start with: What NotchLive tools can you use? Then try Search my NotchLive meetings for the decision about launch timing. For the active Timeline, start captions in NotchLive and ask the client to follow each new finalized caption.

Translate one live stream into several languages

With active meeting captions enabled, ask: Follow my active NotchLive meeting. Translate every new finalized caption into Spanish, Japanese, Turkish, and Arabic. Keep the speaker label and continue until I tell you to stop. Translation quality, language coverage, formatting, and latency depend on the connected model.

Know the boundary

Setup alone does not grant access while the main switch is off. The helper may initialize, but data calls return access_disabled and no meeting content. NotchLive exposes read-only tools and never sends a request by itself, but the connected host may send the text you request to its chosen model provider. Ollama and Qwen need an MCP-capable agent harness. See the full NotchLive MCP guide for the contract, host setup, data boundaries, and live workflows.

Troubleshooting

NotchLive asks to finish setup

Use the Finish Setup button. It opens the right setup step for missing app audio permission or a missing speech model.

NotchLive is missing from Screen Recording

Launch NotchLive once, open System Settings > Privacy & Security > Screen Recording, and add NotchLive from Applications if macOS does not list it. The app also shows a drag-to-add helper during setup.

Captions are inaccurate

Try Medium, Large Turbo Compact, or Large Turbo for accents, dense speech, or technical vocabulary. If you mostly speak one language, set Microphone Language instead of Auto. Also check that the selected audio source matches what you want NotchLive to caption.

The Mac runs hot

Large models use more CPU, memory, and disk space. Switch back to Small or Base if you prefer lower resource use.

Audio does not return after the Mac wakes

NotchLive 1.9.2 rebuilds system-audio and microphone capture after macOS sleep. If recovery cannot finish, use Retry audio capture. Permission failures require restoring the requested macOS permission before retrying.

A speech model needs repair

Use Repair when NotchLive reports a missing or changed model file. The replacement is verified before use, and an interrupted repair keeps the model in a recoverable state.

Your MCP host cannot see NotchLive

For Codex or Claude Desktop, make sure the app was installed before you clicked Set Up, then run Set Up again and reopen the client. For another host, copy the current standard stdio configuration from NotchLive Settings. Turn on Allow AI apps to access NotchLive; live Timeline requests also require the separate Include active meeting captions switch and an active caption session.