Starting your first dictation
Last updated: October 7, 2026
Available on: Mac, Windows, iOS and Android
Flow turns speech into formatted text in the field where you're writing. For your first dictation, try a short sentence in a note or message after completing Flow's setup and required permissions.
Mac and Windows
Keyboard | Push to talk | Hands-free mode |
Mac | Fn | Fn + Space |
Mac without an Apple Fn key | Ctrl + Opt | Ctrl + Opt + Space |
Windows | Ctrl + Win | Ctrl + Win + Space |
These are the defaults; onboarding applies the whole backup set on Macs without an Apple Fn key. Change them in Settings → General → Shortcuts → Change.
Open the app where you want to write and click a text field so the cursor is visible.
Press and hold your Push to talk shortcut.
Speak while holding the shortcut.
Release the shortcut and wait for your formatted text to appear.
Insertion targets the text field focused when the text is inserted. To cancel without inserting, press Esc or your configured cancel shortcut; hands-free recording also has a cancel button.
Hands-free and click-to-start alternatives
Hands-free recording continues without holding a key.
Press your Hands-free mode shortcut, or double-press Push to talk within half a second.
Speak your sentence.
Press the same shortcut again to stop and insert the text; a third quick press after a double-press cancels.
You can also click the Flow Bar to start, keeping the intended text field selected; recording continues after the click, with stop and cancel controls. Clicking the bar again stops it; clicking within half a second of starting cancels instead.
A Push to talk tap released in under half a second does not paste. Flow briefly waits for another press, then cancels and shows the hold-shortcut tip; start again from there.
iPhone
Note: Full Access and microphone permission are separate. If the keyboard shows “Almost there!” because Full Access is missing, tap Finish setup to open Settings. Microphone access is requested on your first dictation attempt; if denied, use Go to Settings to enable it.
Open Flow, sign in and finish onboarding.
Enable the Flow keyboard and its Full Access permission if prompted.
Tap a text field in a note, document or message and switch to the Flow keyboard.
Tap the microphone button, or Start if shown.
Allow microphone access if iOS asks, then speak naturally.
Tap the check button to finish, and wait for the formatted text to appear.
Start can open Flow and begin recording there instead of recording in place. The X cancels without inserting text; during processing, tapping the progress indicator also cancels. Incoming calls and connecting or disconnecting headphones or Bluetooth audio stop recording automatically.
Flow switches to another keyboard in numeric, decimal, phone and email-type fields. If you see a signed-out or no-network screen, sign in or reconnect. For keyboard setup, see Set up the Flow keyboard on iPhone.
Android
Note: Onboarding must be complete before the bubble appears. Its battery-optimization step is skippable: battery setup includes the standard Android exemption on every device, including Pixel. Other manufacturers also require their additional battery or autostart settings; Pixel needs no additional manufacturer setup. Follow Android setup.
Complete onboarding with Microphone, Accessibility Service and Display over other apps enabled.
Tap a text field in another app with the on-screen keyboard visible.
Tap the Flow Bubble at the screen edge.
Speak your sentence.
Tap the checkmark (Done) to finish; text is inserted automatically after transcription.
Alternatively, hold the bubble while speaking and release to submit. Offline devices cannot start dictation, but losing connection during a hold does not prevent release from submitting it.
Tap-to-dictate has an X during recording and transcription; hold-to-dictate has no cancel control during the hold, and it appears about three seconds into transcription. Use the action buttons, not the bubble body, during recording or error recovery.
The bubble hides when the keyboard is dismissed, and restricted banking apps and password, numeric and phone fields suppress it, with narrow exceptions for browser address bars and notification replies.
If the bubble is missing or cannot insert text
Open Flow to see the permission or recovery card for the current problem.
Permissions: use Allow Microphone or Go to Settings to restore the missing permission.
Snoozed: select End snooze now, or shake the unlocked phone while inside a text box.
Flow Bubble Disabled: select Turn Back On; this card appears when permissions are granted but the bubble is turned off.
Flow isn't responding: select Restart Flow when shown, which can take up to about a minute to appear; accessibility permission can remain enabled even when Flow stops responding.
If a separate recovery card appears while Settings already shows accessibility enabled:
Complete the manufacturer's autostart instructions first if the card directs you there.
Turn Flow accessibility off in Settings.
Turn Flow accessibility back on.
Return to a text field with the keyboard showing and try a short dictation. If the bubble still fails, see Flow Bubble troubleshooting.
Check the result
Read the inserted text before sending it. Flow formats lists and paragraphs and handles genuine spoken corrections by keeping the replacement and removing the original wording and correction cue. An explicit cancellation such as “scratch that” removes the retracted wording and cancellation cue. A cue such as “actually” alone is insufficient: ambiguous corrections, unfinished fragments and “never mind” used as part of the message remain.
Warning: On desktop, Flow briefly uses the clipboard for insertion, then restores your previous copy about half a second after the target app reads the text — about three seconds for Outlook (classic and new), Word and WhatsApp on Windows, and about five seconds for recognized remote-desktop apps. Until restoration finishes, Ctrl + V in any app pastes the dictation again.
On Mac, a new copy made during restoration can be overwritten, so wait before copying something else; Windows preserves a new copy. If insertion never reaches an app, Flow leaves the transcript on the clipboard for manual pasting instead of restoring the previous copy.
If dictation does not start or finish
Desktop microphone problems
If no audio is captured, Flow ends without inserting text and shows a microphone, closed-lid or hold-shortcut message.
Message | Action |
Windows: “Microphone access is disabled”; Mac: a microphone permission message. | Follow it to enable microphone access: Windows opens microphone privacy settings, Mac opens Privacy & Security → Microphone. |
Selected microphone is unavailable | Check its connection or select Choose Microphone. |
No microphone detected | Connect a microphone or select Open Sound Settings. |
Microphone unavailable | Close other apps using the microphone; this appears after a repeated access failure, not the first one. |
Unable to access mic | Select Restart App. |
Try a short dictation after the action. If text appears, stop troubleshooting. If the microphone disconnected during recording, select Insert on its message to insert text captured before the disconnect.
Text was not inserted
Desktop: when the failure message displays a transcript, it also copies that text to the clipboard and offers Copy; this replaces your previous copy rather than restoring it. If the message is suppressed, your clipboard is left alone.
Click the intended text field.
Paste with Cmd + V on Mac or Ctrl + V on Windows when the transcript is on the clipboard.
If there is no transcript text, the generic paste prompt names Flow's paste-last-transcript shortcut: Ctrl + Cmd + V on Mac or Alt + Shift + Z on Windows, unless customized. On Windows, “Update admin settings to paste!” directs you to run Flow as Administrator.
Android: Paste retries insertion of the retained transcript, not whatever you most recently copied; the transcript is also copied to the clipboard as a fallback.
Tap Paste on the bubble.
Paste manually if Flow says “Failed to paste text. It is still on your clipboard.”
Check that text appears in the intended field. Tapping the Wispr icon after an insertion failure dismisses the error and keeps the clipboard text. For further help, see Fix text not pasting after dictation.
Processing or transcription did not finish
A dictation started right away either begins after the previous recording finishes or is dismissed, so start it again if it does not begin. On desktop, the “Flow was processing your last transcript” card means the new dictation did not start: wait for processing to finish, then start again.
Dismissing a notification alone does not cancel transcription. Use the card's Cancel button or the cancellation shortcut shown — your configured shortcut, falling back to Esc, which clears the notification and cancels the ongoing work.
For network or transcription failure, check your connection and use Retry or History when offered. Desktop's “Is your internet connected?” notice has no buttons; Retry and Open History are on the separate transcription-error card. Audio is saved for retry when storage settings allow.
Android transcription failures are saved in the app, and Retry uses the captured audio when available; tapping the Wispr icon instead starts a new dictation. Check that the recovered text appears before starting over.
Limits and saved dictations
Follow the weekly word allowance shown in your account. On iPhone, “No words left” blocks recording and offers an upgrade or eligible student-plan option.
Session caps are separate and an upgrade does not raise them. Mac and Windows warn at 19 minutes and stop at 20 minutes with “Flow supports up to 20 minutes per session.” Select Recover text to insert the captured text. Android warns at about 3 minutes 59 seconds, then stops and submits at about 4 minutes 59 seconds.
iOS offers Keep stored, Auto-delete after 24 hours and Never store; the setting applies across your devices.
Important: Past dictations appear in Recent activity or History when your storage settings allow. Cancelling does not guarantee recoverable audio: desktop cancellations under half a second are discarded, storage failures can lose audio, and retention settings can delete saved dictations.
Never store prevents saving, and an organization's Never store policy locks the setting on desktop and iOS — on iOS it also suppresses the failed-dictation retry nudge. Desktop asks for confirmation before stricter retention settings delete history, and removes audio — not transcripts — from recordings older than 14 days, even when stored normally.
Still stuck?
Contact support if dictation still cannot start or text cannot be inserted after the matching recovery steps. Include your platform, device, exact message and the steps you tried.