Docs

Voice.

Some days typing is one step too many. Speak to Kata and have replies read back to you. On Kata Plus and Kata Unlimited the replies come in a Kata voice. One of them is an AI voice made from the voice of Mégane, Kata's founder. This page covers the whole voice suite.

Speak to Kata

  1. Tap the microphone in the chat.
  2. Say it like you would to a friend: "big glass of water and 20 minutes on the bike".
  3. Your words appear as text and send automatically: the message lands like a typed one, and the reply comes with lock-in cards as usual.

People use it while cooking, walking, driving or with wet hands, in casual, real-world speech. On iPhone, Apple's speech recognition turns what you say into text; on Android, your phone's speech recognizer does, which on most phones is Google's and may process the audio on Google's servers; on the web, your browser's speech recognition does. Kata receives that text, never a recording: see Your data. For a full hands-free conversation, where Kata speaks back and you go turn by turn, see Talk to Kata.

Have replies read aloud

On Kata Plus and Kata Unlimited, any reply from Kata can be read to you: tap the speaker icon on a message. Handy mid-workout or when your eyes are busy. Every reply is read in a Kata voice, described next. On Kata Free, the speaker icon shows a calm note that Kata voices come with Kata Plus and Kata Unlimited, and the reply is always there to read.

Kata voices

On Kata Plus and Kata Unlimited, Kata reads replies in a Kata voice: natural, and native to the language of the reply. Rather than one voice stretched across every language, Kata offers voices made for each: German, French, Italian, Spanish, Swedish and English. Each language has its own default voice: Mégane in English, Annika in German, Élodie in French, Chiara in Italian, Carmen in Spanish and Maja in Swedish.

Pick who reads to you in Settings, General, under Kata voices: one list of named voices, grouped by language, each speaking its own. Choose a voice in another language and Kata offers to switch the whole app to it. Numbers stay listenable too: the voice reads units as words, so 80 kg arrives as eighty kilograms. The list shows only the voices that are available, and if the one you picked is ever unavailable, Mégane reads instead. If the Kata voice cannot be reached for a moment, a calm note says so and the reply stays on screen to read.

Mégane's voice, made with AI

For English, the default Kata voice is Mégane: an AI voice made from the voice of Kata's founder, with her consent. It is offered first and marked as the founder voice. Kata writes every word and ElevenLabs speaks it. Mégane does not record, write or send any of it. Because the voice engine is multilingual, the same voice can also read the moments below in your language, with a light accent outside English. You are always free to pick another voice instead.

Voice notes in Mégane's voice

Now and then Kata checks in with a short voice note in Mégane's voice: recognition when you hit a real milestone, like passing thirty days off a habit or reaching a weight goal, and a nudge when you have been away for a while. Kata writes it from what actually happened, never a generic "great job".

  • Where it appears: Kata introduces it in the chat, labeled "Kata AI in Mégane's voice", and it plays as a voice note you can replay any time. If you picked another Kata voice, that voice reads it.
  • How often: rarely, by design. A note like this is special because it is not frequent, and a real win is always chosen over a nag.
  • Your call: it is on by default and you can turn it off in Settings, Notifications, under Voice notes in Mégane's voice, like any other category. Quiet hours are respected, so a note never lands at 3 am.
  • From the team: now and then a voice note from Kata's team can arrive in your notification inbox, with a play button and sometimes a link to what it talks about.

Every voice message plays in a proper player: the real shape of the audio drawn as a wave you can tap or drag to any point, pause and pick up where you left off, replay any time. The voice notes in the chat and the team notes in your inbox share the same one.

The Weekly Kata and Sensei letters, read aloud

The Weekly Kata review and the Sensei letters carry a play button so you can hear them read aloud in Mégane's voice, or in the Kata voice you picked, rather than only reading them. The reading stays faithful to what is on screen. On Kata Free the play button shows a calm note that Kata voices come with Kata Plus and Kata Unlimited, and when the Kata voice cannot be reached for a moment, a note says so. Either way the text is always there to read.

Talk to Kata

Talk to Kata is the hands-free conversation: you speak, and Kata replies out loud in your Kata voice, back and forth, like talking to a person. It has its own page: Talk to Kata.

What voice and Talk cost by plan

A voice interaction is one reply spoken by a Kata voice, or one spoken turn in Talk. Typing and replaying something you already heard never count. Kata voices and Talk come with Kata Plus and Kata Unlimited, and Kata Free has none of them. From 16 November every account gets 30 days of Kata Plus, so you hear them from the start.

PlanKata voices and TalkIn practice
Kata FreeNone.Every reply is there to read, and text chat is untouched. Kata voices and Talk arrive with Kata Plus.
Kata PlusA generous monthly amount.60 Kata voice or Talk interactions each calendar month, Talk included.
Kata UnlimitedThe most, on fair use.400 a month: a high monthly ceiling rather than truly infinite, so Kata voices always stay sustainable.

If you reach the monthly amount, Kata voices and Talk rest for the remainder of the month with a note, replies stay on screen to read, and everything resets at the start of the next month. This is a separate allowance from the fair use of your messages to Kata.

What reaches whom

When you speak to Kata, Apple's speech recognition on iPhone, your phone's speech recognizer on Android (Google's on most phones, which may process the audio on Google's servers) or your browser's on the web turns it into text, and only that text reaches Kata. For a Kata voice, only the reply text to be spoken is sent to ElevenLabs, our voice processor, to be voiced in the moment: the text is turned into audio and not stored there, and nothing new is kept on our servers beyond the aggregate usage count that meters the allowance above. Kata never places phone calls: voice happens only inside the app, when you open it. See Your data.