Kotonoha Support
This page covers the first macOS release. Windows, iOS, and Android are planned for later; platform-specific guidance will be added when each version is available.
What Kotonoha is
It turns what you say into text, where you are already typing. Recognition and formatting both run entirely on your device. Neither your voice nor the text it becomes leaves the machine. (Using it does require signing in — see "Does it work offline?" below.)
Using it (macOS)
- Open Kotonoha and choose a comfortable key and trigger mode in General settings
- Start listening with the key you configured, then speak normally
- Release it, and the shaped text lands at your cursor
You can hold a key while speaking or choose a mode that toggles listening on and off.
Three permissions are needed first
| Permission | What for |
|---|---|
| Microphone | To hear you |
| Accessibility | To put text at your cursor |
| Input Monitoring | To notice the hotkey being pressed |
macOS does not let Accessibility and Input Monitoring be granted from the dialog itself. Use "Open System Settings" and turn the switch on yourself. After turning them on, quit the app and start it again.
Frequently asked
Is what I say sent anywhere?
No. Audio and text are processed on your device only, and are never written to a file. See the Privacy Policy.
Does it work offline?
Yes. Once you have signed in, dictation keeps working.
Recognition and formatting both complete on the device, so nothing on the dictation path needs the network. The same is true on a day our server is down: the only thing that stops is syncing your settings.
The network is used in four situations only: signing in, syncing settings and your dictionary, downloading a speech model when you choose to, and checking at launch whether a newer version exists. None of them carries your voice or your text, and the last one sends nothing at all — it reads one file and compares versions on your machine. You can switch that check off in Settings ▸ Help.
Why does it need a sign-in?
So your settings and dictionary can travel between your devices, and so a purchase carries across them.
What we hold is an email address, your settings and dictionary if you switch syncing on, and whether you have purchased. No password — sign-in is Apple and Google only. Your voice, your transcripts and your history do not pass through it.
Syncing is off by default. Until you turn it on, your dictionary stays on the device — it can contain people's names, so it is your call. See §1 of the Privacy Policy.
It says "press ⌘V" instead of inserting at my cursor
Either Accessibility is not granted, or the field is a secure input field (a password box). For the first, turn Kotonoha on under System Settings ▸ Privacy & Security ▸ Accessibility and restart the app.
Your text is on the clipboard, so ⌘V will paste it — nothing is lost.
The hotkey does nothing
- Open Kotonoha and check it is not paused
- Check Input Monitoring (System Settings ▸ Privacy & Security ▸ Input Monitoring)
- Check no other app has claimed the same key
I do not want a history kept
Settings ▸ General ▸ turn off "save history". With it off, no text and no transcript is recorded at all.
Proper nouns come out wrong
Add them to the dictionary. Settings ▸ Dictionary lets you register "what it sounds like" and "how it should be written", and the replacement is then reliable.
Speaking English gives garbled output
The recognition language is currently fixed when the app starts. Change the language in Settings ▸ General and restart the app. (Automatic language detection is in development.)
Reporting a problem
Email: support@kotonoha.page
Three things make the cause much easier to find:
- What you did (e.g. pressed your configured key and said "tomorrow's meeting")
- What happened (e.g. nothing was inserted)
- What you expected
You do not need to send what you actually dictated.
Where the log is
~/Library/Application Support/app.kotonoha.desktop/voiceinput.log
What you dictated is never written to this log (mechanically checked), so it is safe to attach.
Requirements
| Chip | Apple Silicon (M1 or later) is required. Intel Macs are not supported |
| macOS | 26 or later recommended. The built-in engine is used — 0 GB to download |