Where your voice goes

What leaves your Mac

This is the complete list.

Your audio

Stays on your Mac. NVIDIA's Parakeet model transcribes it there.

Your text, with AI Cleanup off (the default)

Stays on your Mac.

Your text, with Ollama

Stays on your Mac while Ollama runs there (the default). Point Ollama at another machine and the text goes to that machine.

Your text, with Gemini or another cloud provider

Goes to that provider, on your own API key. Off by default.

On launch

One version check to spokenkey.com.

First launch

A one-time download of the speech model, about 500 MB, from Hugging Face.

Analytics and crash reporting

None in the app.

The version check is a plain request to spokenkey.com/api/version once per launch. It has no body and no custom headers, and the app doesn't add your text, audio, or an account to it. If you're offline, it fails silently.

SpokenKey has no analytics, ad tracking, or third-party crash reporter. The app has no code that uploads your history or diagnostics.

If you turn on a cloud provider

AI Cleanup is off by default. The choices are Off, Local (Ollama), Gemini, and a custom endpoint that speaks the OpenAI format. With Gemini or a custom endpoint, SpokenKey sends the transcribed text, never the audio, to the provider you picked. It uses your API key, which you supply. That provider's own terms and privacy policy then apply.

With Ollama, the text stays on your Mac as long as Ollama runs there. The Ollama address is editable in Advanced settings, so if you point it at another machine, the text goes to that machine.

What stays on your Mac

• History. Every dictation, with its raw and cleaned text, mode, and timing, is saved in a local database. There is no sync. You can search it and delete entries.
• Recent recordings. SpokenKey keeps a copy of each recording in a recovery folder so a failed transcription doesn't cost you your words. It keeps the latest 20 and deletes any older than 24 hours when the app launches.
• Diagnostics. On by default, stored on your Mac, and never uploaded automatically. They can include correction notes and provider error details. Sharing is manual: Settings, then Diagnostics, can email a sanitized copy to support or save it to your Desktop. That copy omits dictated or corrected text, document contents, filenames, endpoints, provider messages, and API keys. You can turn diagnostics off in the same place.
• API keys. Stored in the macOS Keychain, and sent only to the provider they belong to.
• Your speech model. Downloaded once, then kept on disk so transcription works offline.

This website

spokenkey.com uses Vercel Analytics to count visits to the site's pages. That measurement exists on the website only. Nothing inside the app reports to it or to any other analytics service.

Purchases go through Polar. Its privacy policy covers your payment and receipt details.

Recording consent

SpokenKey's setup reminds you that you are responsible for making sure recording audio where you are complies with applicable laws, including any consent requirements.

Questions

Email hello@spokenkey.com.