Voice input
Tap the microphone and speak — your words are transcribed into the chat box in any language, ready to review before you send.
Sometimes it's faster to say it than type it — especially on your phone, or when you're logging a long journal entry. Tap the microphone, talk, and Kaizendex turns your speech into text right in the box.
How to use it
The mic sits in the bottom-left of the input, next to the + button (in Your log) or the send button (in an AI agent chat).
- Tap the microphone. It turns red and pulses — Kaizendex is listening.
- Speak naturally — your words appear right in the box as you say them. Each time you pause, the phrase settles into place.
- Tap the mic again to stop — any last words land too.
- Read it over, edit anything if you like, then press ↵ to log or send.
Voice input never sends on its own — it only fills the box. You're always the one who hits enter, so you can fix a word or add to it first.
It works the same in both chat boxes:
- Your log — dictate what you did ("ran 5k and had two coffees") and the logging agent files it for you.
- An AI agent chat — dictate a message to The Navy Seal or Coach Jack.
You can also mix typing and talking: dictate a sentence, type a correction, dictate some more. New speech is added to the end of whatever's already there. (While you're mid-sentence the box briefly locks so the moving words can't be edited out from under you — pause for a beat and it's editable again.)
Speak any language
Out of the box there's nothing to configure. Just talk — Kaizendex detects the language automatically and transcribes it, across dozens of languages. Switch between languages whenever you like (say, Lithuanian for one entry and English for the next); each clip is detected on its own.
Pinning a language
If auto-detect ever mis-hears you — it can occasionally mis-tag a very short clip — you can pin the transcription to one language:
- Open Settings → Profile.
- Under Voice input language, pick your language.
- Save profile.
Pinned means only that language: if you pin Lithuanian, an English recording will come out garbled. So leave it on Auto-detect (any language) — the default — whenever you speak more than one language, and pin only for a stretch where accuracy in a single language matters most. You can switch back to Auto-detect any time.
The first time: allow the microphone
Your browser will ask for permission to use the microphone the first time you tap the button. Choose Allow.
If you accidentally blocked it, the button will show a "Microphone access denied" message. Re-enable it from the little camera/mic icon in your browser's address bar (or your browser's site settings), then tap the mic again.
Your privacy
Recording only happens while the mic is active, and the audio is used only to produce the transcript — it isn't saved or attached to your entry. While you speak, audio streams securely to our transcription service and only the text comes back; nothing is stored. As with the rest of your data, the transcript is yours.
If something goes wrong
- Nothing happened after I stopped. If no speech was detected, you'll see a brief "No speech detected" note — tap the mic and try again, a little closer to the microphone.
- "Couldn't start live transcription" / "Live transcription failed". A momentary network or service hiccup. Just tap the mic and try once more — anything already in the box stays there.
- No mic button at all. Your browser may not support in-page recording (some older or locked-down browsers don't). Typing always works.
Next
- Logging an entry — how the log box turns your words into tracked entries
- Chats and AI agents — who you're talking to
- Quick-log — the one-tap alternative when you don't need words at all