What we do with your voice.

When you hold the dictation hotkey, your microphone records into a temporary file on your device. That file is sent over HTTPS to the transcription model you picked — either OpenAI's GPT-4o transcribe or Google Gemini — through Buna's server-side proxy. Once a transcript comes back, the audio is deleted from the proxy and from your device. We do not store audio.

Long recordings (over 2 MB) are uploaded to your private path in Buna's Cloud Storage bucket so we can stream them to the model. They are deleted automatically as soon as transcription succeeds.

Each successful dictation creates a row in your private history collection (signed-in users only). You can delete any row from History. Transcripts sync across your signed-in devices and are stored in Google Firestore in the United States.

We use Sign in with Apple. We store your Apple-issued user identifier, the display name you choose, and (if you didn't hide it) your email. We don't see your Apple ID password.

We count seconds of audio you transcribe each month, so we can enforce the free tier and bill paid plans. The counter lives in your Firestore document — it's a number per month, nothing else.

None by default. We don't embed Google Analytics, Mixpanel, Amplitude, or any other product analytics SDK. If you opt into crash reporting, only the stack trace and OS version are sent — personally-identifying data is stripped.

We do not train on your audio or your transcripts. OpenAI and Google have their own zero-retention API agreements which we sign on your behalf — your dictations are not used to improve their models either.

Email hello@bunavoice.com from the address tied to your Apple ID and we'll delete every row associated with your account within 7 days. You can also delete transcripts one at a time from the History view.

Anything else, just ask: hello@bunavoice.com.