The live transcript in Notetaker
Last updated: October 16, 2026
Available on: Mac and Windows, where Notetaker is enabled for your account. Account and organization restrictions, including HIPAA restrictions, apply. Mac system audio requires macOS Ventura or later.
Use the Transcript tab to follow meeting speech, find earlier words and copy the conversation. Live text and speaker labels can change after the meeting.
Read the live transcript
Live Transcript is on by default in Settings → Notetaker → Transcription. Starting with it off still saves audio for post-meeting processing, and live transcription can be unavailable for a recording even when the setting is on. Turning it off mid-meeting hides the tab; turning it back on does not start live transcription for a meeting that began with it off, and Flow shows a message that the change applies to the next meeting.
Follow: New speech scrolls into view while you are at the bottom and not searching. Scrolling up pauses following; Jump to bottom resumes it.
Labels: Live views can group your microphone and other audio separately as You/Others, or show numbered speakers. Names replace numbered labels when available; after Resume, the same person can receive a new number.
Markers: Sleep, offline gaps, pause/resume, Dictated and audio events can appear. Hover the divider or use a screen reader for pause/resume times. Dictated means the microphone is being used for Flow dictation; system audio continues recording.
Stop: Trailing lines are saved above a pause divider timestamped when you pressed Stop, and remain when you reopen the meeting from the hub or quit Flow.
Search the transcript
Search and copy are disabled until the transcript has content. Search shortcuts work only with the Transcript tab active and focus outside text fields.
Open search with Cmd+F on Mac, Ctrl+F on Windows, or /.
Enter at least two characters.
Navigate with Enter/↓ for next or Shift+Enter/↑ for previous, or the search arrows; matches wrap around and the transcript scrolls to the selected match.
Search ignores case, matches partial words and treats punctuation literally, with no case-sensitive or whole-word setting. It searches speech, not markers; translated rows use the displayed translation, not the original speech.
During recording, search uses the text present when the search started — clear and rerun the query to include newer speech or corrections. Escape clears and closes search, allowing following again when you are at the bottom. After recording stops, search uses the current transcript.
Understand speaker names
Automatic naming requires the attribute names setting, on by default, and naming availability for the recording; turning it off leaves numbered labels. The supported meeting app must be using your microphone, and on Mac, Wispr Flow also needs Accessibility permission. Browser meetings are matched to the call's own tab, so several calls of the same service open at once can leave speakers numbered. Names are not immediate — Flow can detect the call and begin naming later — and unsupported layouts can leave numbered speakers.
Meeting app | Naming requirements |
Google Meet | Chrome (including Beta/Dev/Canary), Edge or Brave; Chromium or Arc on Mac; Safari unsupported. |
Zoom | Native app with readable meeting windows; on Mac, Accessibility permission for Zoom; on Windows, an English interface. |
Microsoft Teams | Native app on Mac and Windows; Chrome, Edge or Brave on Mac also support naming. |
Slack huddles | Desktop app on Mac or Windows. |
Webex and Discord | No automatic naming; speakers remain numbered. |
Google Meet and Arc on Mac
Keep either the main meeting window or picture-in-picture window visible; each counts independently, and Flow prefers picture-in-picture when both are readable. Full coverage by other applications blocks that window, but Flow's floating panels do not. Only visible participant tiles provide names, and page-layout changes can affect detection. Presenting does not stop naming; the You are presenting tile is ignored.
Flow keeps following the same named Meet call as you switch between the main window and picture-in-picture, and while you start screen sharing or recording. With several Meet calls open, start from the main Meet window so Flow can tell which call to follow. With neither window visible, speech stays unnamed; uncover or restore a window to resume naming, though a brief unnamed interval can remain.
In Arc, naming covers only calls at meet.google.com that Arc is using your microphone for; Teams, Zoom and Slack calls in Arc remain numbered, and Arc on Windows is unsupported. An untitled Arc picture-in-picture window alone cannot provide names. If names are missing while Dia is open, quit Dia and try again: it can prevent Flow from recognizing an Arc call. Participants sharing a display name can be treated as one speaker, and overlapping speech may have no single name.
Teams window restrictions
The native app supports a main meeting window or compact call window; on Windows, the latter is Meeting compact view. Browser-hosted Teams has no compact call window.
On Mac, a main window that is hidden, minimized, on another Space or fully covered cannot supply names. Flow can use the compact window instead, including when covered, but not when hidden or minimized. Restoring a window can leave a brief unnamed interval before names resume.
On Mac, the native Teams app is named from participant tiles, the participant list and gallery pages in any Teams interface language, with no need to open the People panel. The compact window supports the single active-speaker view and gallery; its self-speaking status text is recognized only in English, which does not restrict participant names to English. Your own speech is identified through your microphone, not a highlighted self tile.
On Mac, Flow names speakers from the tiles currently shown, so a multi-page gallery or two people sharing a display name can still be named. The highlighted speaking tile gets the line; with none highlighted, the speaker is left unnamed. If someone on an off-screen page speaks at the same time as a shown participant, only the shown speaker is named. An incomplete read of the participant list or an empty gallery leaves numbered speakers, and a lobby participant does not establish who is speaking.
Your speech can be labelled with your name even when your meeting tile is not marked as you. A single voice in a microphone-only recording appears as You or Name (You). Flow uses a saved speaker rename before your profile name, and a later transcript improvement preserves that rename. First-name-only profiles can remain numbered, and if another participant uses exactly your name, neither is resolved automatically.
Matching speakers to meeting attendees
Flow matches the participants it sees during the call to the people invited to the meeting, so transcript labels use the attendee's name as you know it. A tile can match an attendee by their full name, a short or first name, a joined form of their name, the exact address on the invitation, or the part of that address before the @ sign. Names from your own contacts can match as well, including attendees who have no name on the calendar invite.
When a tile could belong to more than one attendee — two people named John, for example — Flow leaves the speaker numbered instead of guessing. Names you saved yourself, people you linked to a speaker and manual speaker assignments are kept if the meeting is processed again.
Note: Private names that only exist in your own contacts are used to match speakers for you, and are never shown or saved as the shared speaker label others see.
When every other voice has been identified and exactly one invitee is still unaccounted for, Flow gives the remaining voice that person's name — this commonly completes in-person, microphone-only one-on-ones that would otherwise end with a Speaker 2 label. Flow keeps the numbered label instead when the evidence disagrees, the invitee list is incomplete, a meeting room or a declined invitee is on the invite, or the leftover audio looks like echo or media playback. Because of this, a stray sound such as a notification chime can occasionally pick up a real name.
Post-meeting processing also moves individual lines back to the speaker whose voice clearly matches, correcting a sentence that appeared under the wrong name or under an extra speaker created by echo. Lines where two people talk at once are left as recorded, and the number of speakers does not change.
Warning: Speaker names are not a guarantee of identity. Speech containing more than one voice can sometimes carry your name. Review names before relying on or sharing the transcript.
When live text stops
Audio records locally during reconnection, and live text resumes when connectivity returns. Hovering the Flow bar's meeting pill can show Still listening · Reconnecting… when the Notetaker window is unfocused and the meeting is not wrapping up.
Short interruptions catch up; longer outages can leave an Offline for <duration> — audio was captured and will be transcribed after the meeting marker, and post-meeting processing recovers missing words when saved audio is available. Signing out or a revoked session ends reconnection attempts; signing out saves the audio on this device.
If Flow offers Live transcript unavailable because you are offline at the start, recording has not begun.
Choose Record offline to record audio on this device for later transcription, or Cancel.
Select Generate summary after recording, once online, if offered.
The offline offer works once and expires after a newer start attempt. Failed-start cards say whether audio was saved; dismissing, ignoring or letting the card expire does not delete saved audio. See Offline meeting recording for recovery.
Warning: Discard deletes a fresh recording and its saved audio, returning the weekly recording slot it used, and a discarded recording is not uploaded. Typed notes prevent deletion. On a resumed meeting's short-recording card, Discard only dismisses the card and preserves prior content.
Windows: other people's speech is missing
A recorder banner can show Could not capture system audio, System audio capture stopped or System audio permission denied. A concurrent capture can show A meeting is already recording.
Warning: Your microphone keeps recording after a Windows system-audio failure. Recoverable errors can allow other audio to continue or recover. Otherwise, other people's speech and its live translation are lost for the rest of that recording; start a new recording to capture system audio again.
Mac: permissions or an unexpected stop
If macOS asks for Screen Recording permission, restart Flow after granting it; System Audio Recording Only requires no restart. If that permission request remains unanswered, Turn on becomes available again after 30 seconds, and choosing Don't Allow ends the wait immediately. A device another app is using exclusively cannot be captured.
Leaving a native Zoom or Teams call on Mac stops the recording within a few seconds, so anything said after you click Leave usually does not appear; if system audio such as a video starts playing right after, the stop can take around 20–25 seconds. Meetings on Windows are unaffected by this native auto-stop. Browser meetings such as Google Meet use a separate stop check; Zoom web is an exception.
Use Resume if a recording stops earlier than expected, for example after a quick mute/unmute or an input-device change during a quiet moment. After resuming, Flow allows settling time before another automatic stop. Choosing Keep recording prevents another prompt for the same interruption. Resume requires usable, unexpired audio and remaining recording time.
If your Mac sleeps, capture and live transcription stop at that moment, so the saved audio and transcript match the recorded end time. Waking finishes processing the recording; it does not resume it.
Conference rooms and echoed speech
Listening through speakers rather than a headset lets your microphone pick up the other side of the call, so the live transcript can show the same words twice, once as You and once as Others.
Where echo reduction applies, duplicate microphone lines can stop appearing during the call or be hidden without deleting the saved entries. Post-meeting cleanup depends on sustained, reliably detected echo, usable system audio and a preserved original microphone recording; otherwise the microphone audio is kept. Complete duplicate removal and correct speaker naming are not guaranteed.
Warning: Echo cleanup can also remove microphone speech before an echo begins or between echo bursts. The affected interval varies and is not limited to two seconds. Cleanup does not carry across pauses, resumes or Flow dictation interruptions.
Full storage or another saving error
A write failure stops an offline recording or one started without live transcription, keeping audio already saved. Full storage shows Storage Full; other saving errors show a different error notification.
Warning: A recording started with live transcription online can continue with only its live transcript after a saving failure. Post-meeting improvement is skipped, and speech saved only in audio — including offline gaps — can be lost.
Press Stop and the meeting finishes promptly with a storage or processing failure, keeping the audio recorded up to that point. Free up space and restart Flow, and Flow re-processes that audio on its own. This recovery applies to the retained audio from an offline recording, not audio discarded after a live-transcribed recording's saving failure.
After the meeting
Keep Flow online for processing. Waiting for a connection appears in the floating recorder or the note opened inside the hub — not on hub list rows or the side preview. The meeting stays listed, and uploading resumes a couple of seconds after connectivity returns.
Use that message's Retry control for a stuck upload, including one already in progress; wait 15 seconds between attempts. Retry needs cloud sync on and the recording finished preparing. Busy servers show Servers are busy. Flow will retry at {time}. or Servers are busy. Flow will retry automatically. Other refused retries may show no message. While you are signed out, Flow pauses these uploads instead of attempting them, so time spent signed out never uses up a recording's upload attempts; signing back in starts the pending uploads again right away.
Local audio stays on your computer until the stop has been saved and the audio prepared, so an interrupted stop does not cost you the recording. Reopening Flow finishes interrupted recordings with surviving audio, including audio captured after Resume, and resumes interrupted uploads; missing or empty audio cannot be recovered. Silent or very short recordings can show The meeting was too short to summarize.
Warning: Deleting a meeting permanently removes its note, transcript and audio after you confirm Delete meeting?. If deletion fails, the meeting reappears with an error.
Warning: Local audio remains after transcript processing and is kept for seven days by default — the default resume window. Uploaded audio is kept for seven days from upload by default. See meeting-data retention.
Copy or share a transcript
Click Copy transcript, or press Cmd+C on Mac or Ctrl+C on Windows with the Transcript tab active, nothing selected and focus outside text fields. The whole transcript copies as Speaker: text lines using displayed labels and translations, excluding markers and deleted lines. A manual selection copies only the selection.
My thoughts stays private. Shared summaries can still contain transcript-derived content and speaker names, even under a notes-only organization policy.
Warning: Check copied content before sending it. Invitation emails can include a summary preview — not raw notes or transcripts — and revoking access cannot retract an email already sent.
Sharing restrictions and access changes
The owner's organization policy applies even to recipients in other organizations, and changes apply without restarting. An invitation cannot override an organization-policy denial, and organizations can disable sharing entirely; ask the owner's administrator about access under the current policy.
A shared note's disabled Transcript tab explains whether an invitation or organization policy blocks access. On the shared-note website, even recipients with transcript access must open the note in the Wispr Flow desktop app.
If access is missing, check the account in the web header and use Switch account. There is no Request access button; access-denied screens offer Sign in or Retry.
Tightening link visibility does not remove explicit invitees. If the owner leaves their organization, organization-based access ends, but the owner and explicit invitees keep access.
Still need help?
Use transcript and summary troubleshooting for missing or stuck results. If text is still missing after the relevant recovery steps, contact support with your platform, meeting app and the steps you tried.