Dictate to your computer while pacing the room

Hold the button on your phone, speak, and your text appears in the active document on your computer when you let go.

LibreOffice Writer on Ubuntu with the sentence on the page and the cursor right after it. — The sentence landed in LibreOffice Writer after tapping Deliver while the phone stayed on a desk.

The talk button is on your phone, so you can dictate from across the room.

We have not tested range, walking, or movement. In our tests, the phones stayed on a desk.

The computer needs no microphone, no virtual microphone, and no audio driver, and the desktop app never asks for microphone permissions.

You can check the recognized words on your phone screen while you talk.

In a live words test on Local LAN in Realtime mode with a 19-second synthetic speech sentence across 3 of 3 runs, words on the phone moved for the first 8 to 9 seconds, then froze while the button stayed held down, showing the message "No new words have come back for a while". After release, the full sentence arrived and landed in Notepad in 3 of 3 runs. The phone view shows progress rather than final text. Cloud relay was not tested here.

Phone app while the button is held, showing transcribed words, a timer at 0:06, and a red Release to end button. — The phone view shows progress rather than final text.

Phone app showing a red warning that no new words have come back for a while, with the button held. — To show this message, we played steady noise after the sentence. Recognition did not stall on its own.

What does not work or was not tested

The talk button moves from the desk to your hand, but you must keep holding the button while you speak. Your target window on the computer must stay in front, and your cursor must sit in the text box before you begin. Words do not stream into the document as you speak; words land only after you let go. The phone app needs to stay open on your screen while you talk.
Words land in whichever window sits in front at delivery. In one test, a freshly opened browser window grabbed the foreground, sending a sentence meant for Notepad into the browser.
Password fields on Windows are NOT refused. In a Chrome password field, 2 of 2 sends landed directly in the password field, and the phone showed "Delivered". Check your cursor position before talking.
FlowMic does not replace a mouse, and it cannot click or scroll through your windows. There are no voice commands for switching programs, highlighting text, or moving around the screen.
Desktop versions cover Windows (with a digitally signed installer that may still prompt on first launch), macOS 11+ on Apple silicon with Accessibility permission, and Linux x86-64 on X11 (.deb and portable zip), while phones require Android 7+ or iOS 15+ (the App Store version is unavailable in mainland China and the European Union). Only Windows and Linux paired with an Android tablet have been tested; macOS and iOS were not tested.
Recognition of a real human voice was not measured: our tests checked where text lands and how apps behave using a synthetic computer voice from a second phone beside the tablet microphone. Most errors we saw began in recognition.

More limits of this kind

On 6 November 2026 we ran 5 spoken takes with "Direct send" using a script holding the button, phone app 0.3.118, desktop 0.3.120, and "Local LAN". With a synthetic voice from a second phone's speaker beside the tablet, in all 5, letting go sent the text with no card and no "Deliver" tap. 3 landed: 2 in LibreOffice Writer on Ubuntu under X11 as raw text with no language model, 1 in Windows Command Prompt translated into Spanish by a language model on that computer. The other 2 were refused with the row "Delivered · no focus, not injected · Resend", because a console menu was open on purpose. That makes it a spot check, not a measurement.
A copy made while FlowMic pastes can be lost.
When a console is in Mark mode selecting text, words are lost and nothing appears, while the computer says sent and the phone says "Delivered". That occurred in 3 of 3 on Windows PowerShell 5.1 on a test build with the same Windows code.
In terminal panels in VS Code and Cursor, and console windows started directly by ssh, python or wsl, line breaks are not joined. A line break can run a command there, so send one line at a time.
On Linux in 0.3.120, GNOME Terminal at a bash prompt reached the command line in 5 of 5 sends, and nothing ran. xterm and the terminal panel inside VS Code receive nothing, while the computer row says "Input sent" with "(unconfirmed)" and the phone says "Delivered".
Not tested on Windows 11, for cursor position after insert, or in Word, Outlook, Gmail, Slack, Notion, Google Docs, Overleaf, Teams, Discord, Obsidian, VS Code, and helpdesk tools. Cloud relay was not evaluated during this test set because the test computer was not signed in.

A few common questions

Does the Delivered status mean the words were typed into your window?

The label "Delivered" means the words reached your computer, not that they were typed into a window. The text goes to whichever window is in front when it lands. Look at that window to see whether it arrived.

Can you recover text if a window did not accept it?

The phone and computer keep separate timelines of your sentences. If text reaches the computer but cannot type, the phone timeline shows "Delivered · not injected · Resend" or "Delivered · no focus, not injected · Resend". From the phone timeline, you can tap "Resend", "Copy", or "Copy original text", while the computer timeline offers "Re-inject".

Download Try it

Tested on FlowMic 0.3.120 on the computer and 0.3.118 on the phone, 6 November 2026.

Some counts come from FlowMic 0.3.119 on 5 November 2026 or noted test builds and were not repeated on 0.3.120.