The differentiator
Private on-device dictation for Mac
Most dictation apps send your voice to a server and offer privacy as a mode you can switch on. Sayframe does it the other way round: transcription runs on your Mac, and the network is the exception you choose. This page is the precise version of that claim, including the parts where data does leave, and how to check it yourself.
What runs on your Mac
Every item here is free, works with no account, and needs no connection. Local compute costs us nothing to give away, so we do not charge for it.
Speech to text
whisper.cpp and Apple Speech run on your Mac, with Parakeet available as a faster local option.
AI cleanup
Clean up filler and grammar with a local model, or with your own provider key if you want a hosted one.
Meeting capture
Recording and transcription of a meeting happen on the device, with no bot joining the call.
Transcript Library
Transcripts, search, retry state, and export live on your Mac, where you can delete them.
Modes and per-app rules
Prompts and per-app behavior are local settings, not a profile held on a server.
What leaves your Mac, and exactly when
A privacy page that only lists what stays local is marketing. These are the three paths where data can leave, stated the same way the privacy policy states them.
Sayframe Cloud transcription
Only when you opt in, and only for that recordingSayframe Cloud requires sign-in. The audio is uploaded to our hosted service and a configured speech-to-text processor. Staged audio is deleted after processing. The resulting transcript can remain available to your signed-in devices for up to 24 hours, then its text is removed from our job store.
Cleanup with your own provider key
Only if you configure a keyText you send for cleanup goes to that provider under their terms, not ours. This is the one path where a third party can see your words without Sayframe Cloud being involved, and it is worth knowing before you paste a key in.
Anonymous product analytics
Off by defaultIf you opt in and analytics are configured, Sayframe sends a random installation identifier and coarse product events. Never recordings, never transcript text.
Verify it in about a minute
You should not have to take a privacy claim on faith. Three checks, none of which require trusting us.
- 1Turn off Wi-Fi, press your shortcut, and dictate. Local transcription keeps working, because nothing in that path needs the network.
- 2Watch network activity while you dictate, with Little Snitch or the tools already on your Mac. Local dictation makes no transcription request.
- 3Do not sign in. Sayframe Cloud requires an account, so with no account there is no hosted path to reach by accident.
Why the default matters more than the mode
Several good Mac dictation apps can keep audio on your machine. Spokenly ships a Local Only Mode that makes no network requests, and Wispr Flow has an opt-in Privacy Mode. Both start in the cloud. The difference is not capability, it is which way the switch points when you install the app and forget about it.
If you want the built-in option instead, Apple Dictation is free and already on your Mac, though on older macOS it sends audio to Apple on every phrase and the data flow is not configurable.