Advanced voice-to-text dictation for the Mac, with AI built in.
Hold a key, speak, and polished text lands at your cursor in any app. On-device speech recognition and an AI cleanup pass turn spoken thoughts into finished writing in about a second, without sending audio anywhere.
Dictation should be faster than typing. In practice, the capable tools route your voice through someone else's server, add lag, and stumble on names and technical terms, while the private ones are slow or clumsy. People who think out loud for a living need both speed and privacy.
A lightweight menu-bar app. Hold Fn, speak, release, and the cleaned-up sentence is pasted where your cursor is, in any application. Every dictation is kept and searchable, with its audio and both the raw and polished text, so nothing you said is lost.
Two models run on the Mac itself: a speech model transcribes, and a language model cleans the result using the app you're in, your dictionary and your preferred style. A frontier model is reserved for the prompt-rewrite feature, where quality matters more than milliseconds. The result is cloud-grade output with local-only privacy.
Latency is measured on every dictation and tuned in deliberate steps: cached prompt prefixes, context reads that overlap with transcription, and models kept warm so they're ready before you are. A prompt-quality evaluation suite, run on real dictations, guards the cleanup pass against regressions.
Snippets, a per-app style editor, and a wider release. The faster paths are already measured; shipping them is a matter of priority.
We co-build products with founders and ship AI work for companies. Tell us what you're building.
Book a callOr email hello@gorincollective.com