Privacy Policy
Last updated September 2026.
Local dictation (default)
By default, SpeakenUp transcribes entirely on your device using a local whisper.cpp model. Your voice never leaves your computer, and SpeakenUp has no servers involved in this path at all. Every completed transcript is saved locally, in a SQLite database on your own machine — never uploaded anywhere. You can delete any entry, or set a retention limit, from within the app.
Bring-your-own-key cloud providers
If you configure a third-party provider (OpenAI, Groq, or OpenRouter) with your own API key, your recorded audio is sent directly from your device to that provider for transcription, using your credentials. SpeakenUp doesn't see, store, or relay that audio — it's a direct connection between your app and the vendor you chose, governed by that vendor's own privacy policy.
SpeakenUp Cloud (optional)
If you activate SpeakenUp Cloud, your recorded audio is sent to SpeakenUp's own transcription proxy, which produces your transcript and returns the text. To do this it uses a third-party speech-processing provider, which receives the audio only to transcribe it. The audio itself is never stored or logged by SpeakenUp. SpeakenUp Cloud records only anonymous, aggregate usage metrics (request duration, which model was used, and success/failure status) tied to your account or trial device, used solely to enforce usage limits and monitor service health — never transcript text, never audio content.
Meetings
If you use Meetings (a SpeakenUp Pro feature), SpeakenUp records your microphone and your system audio for the duration of the meeting — system audio can include the voice of other participants on the call, not just you. The recording, its transcript, and its summary are saved locally on your Mac and stay there until you delete the meeting; nothing uploads automatically. If you choose to process a meeting, its audio is sent to SpeakenUp Cloud, which passes it to a third-party speech-processing provider to produce the transcript and to tell the speakers apart. Telling speakers apart only labels who spoke when ("Speaker 1", "Speaker 2"). It doesn't identify anyone, and a speaker gets a name only if you give one. SpeakenUp Cloud deletes the audio and transcript from that provider as soon as it has your result, and the provider's own retention is set to one day at most as a safety net. We have opted out of the provider using your audio to train its models.
The transcript text, and any speaker names you added, is then sent to a third-party language-model provider to write the summary and title. SpeakenUp Cloud doesn't store the audio or transcript content; it keeps only usage metrics, such as the meeting's duration, to enforce your monthly allowance. Both providers process this data only to return your result.
Meetings can capture other people's voices. You're responsible for letting the other participants know a meeting is being recorded.
Account (optional)
SpeakenUp Cloud's free trial starts anonymously, tied to a hardware-derived device identifier — no account required. If you log in for a paid subscription, SpeakenUp stores your email address and a Stripe customer identifier for billing. Payment details themselves are handled entirely by Stripe; SpeakenUp never sees or stores your card information.
What SpeakenUp doesn't do
No advertising, no ad tracking, no analytics on the local or bring-your-own-key path, and we never sell your data to anyone.
Contact
Questions about this policy? Reach out at [email protected].