The live transcript in Notetaker

The live transcript is a running record of your meeting: speech appears in real time, and once the call ends Wispr produces a refined transcript with speaker names. Use it to catch up after joining late, check an exact quote, or copy the record elsewhere.

How it works in Flow

The Transcript tab

• Where it lives: meeting view tabs run Notes · Transcript · Summary. In the meeting window the transcript also sits in a resizable drawer with Transcript and Chat tabs, whose open state and height persist across meetings and restarts. Esc clears focus from an input, then minimizes the drawer.

• Defaults: your first recording opens on the Transcript tab with a "Welcome to Notetaker!" prompt that clears once speech starts; later meetings default to Summary (ended or past) or My Notes (live). When a meeting ends, Notetaker switches to the Summary tab unless no summary is expected.

• Section bar: while recording, a green three-bar voicewave reacts to your mic beside a live M:SS timer; after you stop, the timer becomes speaker count and duration ("3 speakers · 42:10", hidden during a search). Search and copy sit in the bar (disabled while the transcript is empty), and Assign speakers appears once speakers are detected, on meetings you own.

• Consent reminder: at recording start, "Always get consent when transcribing others. Learn more" shows above the composer briefly; "Learn more" opens the Wispr Flow Terms of Service. It's skipped if the Ask Anything drawer stays open all recording, or for unopened background recordings ("Open notepad automatically" off).

Reading the live transcript

• Bubbles: speech appears as chat-style bubbles sized to their content, with a report-flag icon in the right gutter on hover. Mid-sentence speech shows inline in gray, then solidifies once finalized. The live view has no speaker labels — those arrive with the refined transcript.

• Reporting a transcript issue: click the report-flag icon on a bubble, then select Important word transcribed incorrectly, Name transcribed incorrectly, Wrong language, or Other. Submit stays disabled until a category and any required feedback are provided.

• Scrolling: the view auto-follows while you're at the bottom; scrolling up releases it and a Jump to bottom button appears. Past or shared transcripts open at the most recent entry.

• Markers: a Mac sleep inserts a readable marker at the gap ("Your device slept for 1 hr 11 min"); pausing and resuming inserts a gap divider whose hover shows the exact times ("Paused 7:30 PM – Resumed 7:31 PM"). No transcript content is altered.

• Reconnecting after a network blip: an interrupted connection shows a "Still listening · Reconnecting…" indicator and keeps retrying; the live transcript resumes once your network recovers. Your recording, notes, and final summary are not affected.

Live word count on the Flow Bar

• What it shows: hovering over an active meeting in the Flow Bar shows a live word count next to the meeting title, updating as people speak.

• When it's hidden: before anyone has spoken (no false "0 words"), when the live transcript is unavailable, or while Notetaker is reconnecting.

Speaker labels and names

• Refined labels: speakers are labeled "Speaker 1," "Speaker 2," and so on by who speaks first. You appear as "Your Name (You)" — for example, "Jane Doe (You)"; exports show your profile name instead of "You."

• Auto-detected names: with Auto-detect speaker names on (the default, in Settings → Notetaker), Flow attaches real names in the refined transcript when confident, once, when the transcript is first saved.

• How names resolve: candidates come from your calendar invite and the meeting app's participant tiles, deduplicated. A short tile name ("Samm") merges with its fuller calendar version ("Samm Yates"); a space-less calendar name ("Eliolcott") matches its spaced form ("Eli Olcott") only when each part is at least 3 characters, so "Jo Han" and "Johan" stay separate; your own short profile name never absorbs a longer name like "Joshua Kim." Ambiguity — one tile "Samm" with attendees "Samm Yates" and "Samm Turner," or duplicate display names — leaves people unmerged. Manual assignments are never overwritten, "Add a person" adds people manually, and re-refining never deletes a person.

• Owner identification: Flow identifies you as the meeting owner using the meeting app's own "self" signal from your participant tile, not by matching tile names to your profile display name, so you and a participant who shares your display name stay separate speakers.

• 1:1 calendar meetings: for calendar-linked meetings with exactly one other invitee, both you and that person are named automatically — including Slack huddles and calls over laptop speakers, so you'll see real names rather than "Speaker 1 / Speaker 2" even without headphones. The exception is two people sharing a single mic: those speakers stay unlabeled to avoid wrong names.

• Ad-hoc meetings: with no invite or participant tiles, Flow names a speaker from the transcript itself when someone self-introduces or is addressed by name at least twice, with high confidence they're present. Otherwise the label stays "Speaker N" or "Unknown" — Flow never guesses.

• Mic-only meetings: distinct voices stay separate; the tradeoff is one person occasionally splitting into two anonymous speakers. Your profile name is used consistently for your own mic across the transcript and summary, or "Speaker N" if your profile has no name set.

• In-person recordings with multiple voices: when multiple people speak into the same mic during a live (unrefined) recording, speaker mentions in the Summary show as "Speaker N" without a naming popover, rather than being attributed to the recording owner. Speakers you've already named keep their name and color. Full attribution applies once the refined transcript is ready. Remote calls (Zoom, Google Meet), solo recordings, and paused-then-resumed recordings are not affected.

• Google Meet and Zoom: Meet naming needs the Meet window visible in Chrome, Chrome Dev, or Chromium (not Safari); Zoom naming needs the window accessible and macOS accessibility permission for Zoom. Both are rolling out gradually, so some calls still show "Speaker N."

• Speaker chips in summaries: unmatched speaker ordinals still appear as clickable chips in the summary and update to the correct name once the speaker is identified. When the same speaker number arrives from both your mic and system audio simultaneously, that speaker shows as plain "Speaker N" text to avoid attributing a quote to the wrong person. The Flow Summary in the notes panel always matches the Summary tab.

• Manual naming: click Assign speakers, or click any underlined speaker mention in the summary to name it in place (dotted underline = unassigned). Modal rows are pre-filled with the current resolved name (such as "Josh Benson (You)"), the dropdown lists the full roster, the "Same person" tag marks people already on another row, and overtyping a named row renames the person. Renames apply instantly across transcript, summary, action items, previews, and search — the summary is not regenerated. On notes shared with you, names are plain labels and can't be edited.

• Names in summaries and notes: where all speakers are named, every person who spoke appears as a name in both the summary and the notes document, with no leftover "Speaker N" entries. Only people who did not speak appear as plain text.

• Bubble chunking: the refined transcript groups speech into sentence-aligned bubbles, starting a new one only on a speaker change, an audio event, a short pause, two completed sentences, or a long run-on — never mid-word.

• Conference rooms: brief echo dips are smoothed so only sustained changes update live attribution; post-call talk appears in the live transcript shortly afterward.

Search, copy, and export

• Search: press Cmd+F or /, or click the search icon. Matches highlight yellow (current match orange); step with the arrows or Enter / Shift+Enter; "0/0" means no matches. Mid-recording, results freeze to the text that existed when you searched — clear or reopen to include new lines. Esc closes the search, then exits the meeting view.

• Copy: click Copy, or press Cmd+C with nothing selected, to copy the full transcript as rich text (a toast confirms); with a selection, Cmd+C copies just that. Copied text shows resolved speaker names.

• Copy to Markdown: the notepad-header button exports the whole meeting as one Markdown document — meeting info and attendees, Flow Summary, pre-read, notes, action items, chat messages, and the full transcript — with assigned names and your profile name instead of "You."

Chat and "What did I miss?"

While a meeting records, the Chat tab's What did I miss? chip returns an instant recap of the last couple of minutes. In video meetings your own mic lines are labeled "You:"; mic-only meetings keep diarized labels. If nothing new was said it replies "Nothing new since you last looked."; errors show "Something went wrong. Please try again."

Afterward, Ask Anything in the hub answers quote-level questions from the actual transcript text; broader questions ("what meetings did I have this week?") search your meeting notes instead.

After the meeting

• Generation band: a compact band above the composer shows each stage ("Uploading your audio," "Polishing transcript," "Generating notes") with a step counter. Failures show a single Retry button; too-short recordings show "too short to summarize" with no retry.

• Reliability: a closed laptop or dropped connection only delays things — Flow retries when your computer wakes, and the refined transcript reaches all your devices, even with unsynced local edits. Signing in after launching the app immediately pulls down any waiting refined transcripts, without a relaunch.

• After resuming a recording: the existing transcript is kept, with a "New recording since this transcript" notice and an Update Summary button. When the resumed recording's audio covers the full meeting, the summary generates automatically, the button clears, and the refined transcript (with speaker names) replaces the raw live transcript. If a summary is still generating when you resume, Update Summary stays available so you can regenerate once all segments are in. Not updating changes nothing; failures leave the transcript untouched with a Retry option. Too-short recordings update automatically. Manually edited summaries are never silently overwritten.

• Retention: a transcript removed by your retention settings shows an empty state linking to your retention policy; deleted transcripts cannot be restored.

Common issues

Turning the live transcript off

1. Open Settings → Notetaker.

2. Switch off Show live transcript (on by default). A toast confirms.

The Transcript tab hides from your next recording onward; audio is still captured and a refined transcript with speaker names still generates after the meeting.

FAQs

Limitations and notes

• This feature is available on Mac only.

• It works on any plan, including Basic/free, when signed in with Cloud Sync on; accounts with Cloud Sync off (including signed HIPAA BAA accounts) see "Cloud Sync must be enabled to use this feature."

• Speaker-name assignment in the summary is rolling out gradually; existing meetings pick up improved name resolution only after re-refining — there is no automatic backfill.

• Automatic naming for 1:1 meetings applies when the meeting is calendar-linked with exactly one other invitee. Meetings where two people share one mic stay unlabeled.

• Search covers finalized transcript text only, with no case-sensitive or whole-word options.

• A mic drop with the meeting continuing on system audio only shows no in-meeting notification, and no indicator shows which mic Notetaker is using. Device changes switch to the new default almost immediately; an unrecoverable mic falls back to system audio so the meeting keeps going.

• The "Still listening · Reconnecting…" indicator is generic, with no attempt count or time elapsed.

• AI prompt suggestion chips in the chat and hub (such as starter prompts and follow-up suggestions) are rolling out gradually. The What did I miss? chip in the Chat tab is available to all users during a live recording.