NotchLiveFollow every sentence
Keep Focus view on the current line, or search Timeline without stopping new captions.
Meeting transcription software for Mac
A meeting transcript you can trust before you share it.
NotchLive captions the conversation live, separates remote speakers on your Mac, and gives you a review workspace to fix names, turns, and transcript noise before analytics, notes, or exports use the result.
No meeting bot · No cloud transcription workspace · No account required
New in NotchLive 1.9
Stay with the meeting now. Keep the useful parts later.
NotchLive starts as a live caption app, so the transcript helps while people are still speaking. When the meeting ends, the same timeline becomes a reviewable record with local speaker labels, participant corrections, analytics, notes, and exports.
Confirm who said what
Name participants, fix mixed turns, merge duplicate labels, and leave uncertain speech visibly unresolved.
Work from the version you trust
Copy, export, measure, or generate notes from the reviewed transcript instead of the first machine pass.
Participant review
Fix the transcript where the model needs you.
Automatic speaker labels are a useful first pass, not a final identity claim. NotchLive keeps the uncertain parts editable and makes every correction reversible.
- 01
Name the people you know
Replace Speaker 1 with the right name for this meeting, or accept a useful suggestion.
- 02
Merge the same voice
Combine duplicate labels when one participant was split into several anonymous clusters.
- 03
Move a line to the right person
Reassign a full utterance, add a participant, or keep it marked unknown when the evidence is weak.
- 04
Split a mixed turn
Select the words spoken by another person and move only that part to the correct participant.
- 05
Remove transcript noise
Hide an irrelevant caption from the Clean transcript while the captured Raw version remains available.
- 06
Undo without fear
Reverse the latest correction and regenerate clean output so old names do not survive by accident.
Transcript control
Keep the source. Share the clean version.
NotchLive preserves the captured transcript as Raw and builds a Clean version around your corrections. You can improve what people read without losing the source you may need to inspect later.
The captured transcript
Return to the wording and turns retained by the speech-recognition pipeline.
The readable transcript
Read, copy, and export a clearer view while Raw remains available for comparison.
The trusted meeting record
Use confirmed participants and corrected turns for analytics, notes, copy, and export.
Local speaker diarization
Separate remote voices without sending the meeting away.
On Apple silicon, NotchLive keeps your microphone and app audio in separate lanes. Your microphone stays labeled You. SpeakerKit groups remote voices into anonymous, meeting-local participants, then NotchLive aligns those labels with the locally transcribed timeline.
See how local speaker diarization works, including voice embeddings, clustering, final review, and the limits of automatic labels.
Your microphone
Stays assigned to You instead of becoming another anonymous voice.
Remote app audio
Becomes meeting-local speaker labels through SpeakerKit on your Mac.
Your corrections
Become the speaker names used by analytics, notes, copy, and exports.
Meeting context
Choose the meeting shape you already know.
A one-to-one call should not be treated like a ten-person meeting. NotchLive lets you give the local speaker model simple context before it starts guessing.
Auto
Let SpeakerKit estimate the number of remote voices when you do not know who will join.
1:1
Tell NotchLive there is one remote voice so a single person is less likely to fragment into several labels.
Group
Enter the total number of people, including yourself, so the final pass can use the expected remote count.
From evidence to action
Measure the conversation. Then turn it into action.
Your reviewed transcript powers two useful layers. Meeting Analytics shows what happened in the conversation. Meeting Intelligence turns that record into decisions, questions, actions, topics, and a summary on supported Macs.
See how the conversation moved
- MeasuredWords, turns, speaking time, conversation share, participant metrics, and attribution coverage.
- Built fromCalculations from the finalized or reviewed transcript.
Keep the outcome of the meeting
- GeneratedCurrent focus, key points, questions, actions, entities, decisions, and summaries.
- Built withOptional Apple Intelligence generation on a supported Mac.
Your meeting stays a local working record.
Speech recognition, remote-speaker attribution, participant corrections, and Meeting Analytics run on your Mac. NotchLive does not create a cloud meeting workspace or use your audio to train a NotchLive model.
Automatic labels are anonymous and meeting-local. NotchLive does not silently build persistent voiceprints or recognize the same person across meetings. Voice centroids used to keep labels stable are held in memory for the active run and discarded with it.
Use your own data
Export the record, move the library, or query it locally.
NotchLive keeps saved sessions and Voice Notes inspectable. Pro users can export transcript formats for everyday work and explicitly connect supported local AI clients through read-only MCP. Settings can also create a versioned data bundle for backup or transfer.
Own the record
- Everyday exportCopy, TXT, Markdown, and JSON use reviewed participant names while keeping internal speaker IDs out of ordinary text.
- Data bundleMove sessions, notes, participant review, and rebuildable analytics without an account or opaque cloud migration.
Choose when AI can read it
- Off by defaultA configured client receives no meeting data until you enable AI Connections in NotchLive Settings.
- Saved libraryCodex or Claude Desktop can search and read reviewed meetings, analytics, and Voice Notes through local read-only MCP.
Transcribe Zoom, Google Meet, Teams, and more.
NotchLive works with capturable Mac app audio, so the meeting platform does not need a NotchLive integration or an invited bot.
Zoom
Google Meet
Microsoft Teams
Browser calls
Webinars
Lectures
Meeting transcription software FAQ
What is new in NotchLive 1.9 for meeting transcription?
NotchLive 1.9 adds local remote-speaker attribution on Apple silicon, Auto/1:1/Group meeting context, participant review, Meeting Analytics, Raw and Clean transcript views, transcript cleanup controls, portable data bundles, and opt-in read-only local MCP access.
What is the best meeting transcription software for Mac?
NotchLive is built for Mac users who want private live captions, local transcription, meeting-local speaker labels, participant review, and conversation analytics without inviting a meeting bot or uploading audio to a NotchLive transcription service.
What is speaker diarization?
Speaker diarization is the process of determining who spoke when. It separates speech into anonymous speaker clusters and aligns those clusters with transcript words. It does not automatically know a person's real name.
Can NotchLive identify different speakers in a meeting?
On Apple silicon, NotchLive can separate remote voices into anonymous, meeting-local speaker labels while microphone speech remains assigned to You. You can review and name participants before relying on the final transcript or analytics.
Can I correct speaker labels after a meeting?
Yes. In a recorded session, you can rename participants, merge duplicate labels, reassign an utterance, mark speech as unknown, create a new participant, split an utterance at word boundaries, and undo the latest edit.
Does NotchLive transcribe Zoom, Google Meet, and Microsoft Teams without a bot?
Yes. NotchLive captions capturable app audio from Zoom, Google Meet, Microsoft Teams, and other Mac apps. It runs as a Mac app and does not join the participant list.
Does meeting transcription run locally?
Yes. NotchLive uses local WhisperKit speech recognition and bundles the selected SpeakerKit models used for remote-speaker attribution on Apple silicon. NotchLive does not upload meeting audio to a cloud transcription service.
What is the difference between Meeting Analytics and Meeting Intelligence?
Meeting Analytics calculates words, turns, valid-timing speech, conversation share, and speaker-label coverage from the transcript. Meeting Intelligence is the optional Apple Intelligence layer that generates focus, key points, questions, actions, and summaries on supported Macs.
Does speaker attribution require Apple silicon?
Automatic remote-speaker attribution requires an Apple silicon Mac. Live captions continue to work on supported Intel Macs without automatic remote-speaker separation.
Can Codex or Claude Desktop read my NotchLive meetings?
Yes, after you configure NotchLive's local read-only MCP server and enable AI Connections. Supported clients can search saved meetings, read reviewed transcripts, Meeting Analytics, and Voice Notes. A separate off-by-default setting can expose the active Timeline. NotchLive exposes no mutation or network tools through MCP.
Start with a transcript you can review.
Download NotchLive for free live captions and automatic speaker labels on Apple silicon. Upgrade to Pro when you need Meeting Analytics, participant review, saved sessions, exports, translation, Voice Notes, or Meeting Intelligence.