Hold a key. Talk. Your words appear — and they never leave your Mac.
Rhino is a fast, private, local-only dictation app for macOS. On-device speech models, on-device AI cleanup, no account, no telemetry. A private downstream of the MIT-licensed OpenSuperWhisper — see ACKNOWLEDGMENTS.md.
Status: pre-release, building toward a signed, notarized DMG. Requires macOS 14+, Apple Silicon recommended.
- ⌨️ Global shortcut — a key combo or a single modifier (Right ⌥, Fn…), with hold-to-record: hold to speak, release to insert into any app.
- ⚡ Fast — Parakeet on-device with a live preview as you speak.
- 📖 Dictionary — your names and jargon spelled right, every time.
- 🧠 AI cleanup — optional punctuation/casing cleanup on the built-in on-device model (Qwen2.5 1.5B, runs in-process). The raw transcription is kept verbatim if the model misbehaves.
- 🗂 Local history + stats — everything stores to THIS Mac: search past dictations, rerun with a different model, see words dictated and time saved. One toggle turns it off; retention caps in between.
- 📁 Files too — drag audio in, or
Rhino transcribe file.m4afrom the terminal.
Your audio, transcripts, context, and AI cleanup never leave your Mac.
There is deliberately no remote transcription engine and no telemetry. The
only network calls are update checks (signed Sparkle feed) and model
downloads you trigger yourself — enforced by tests/hygiene, which fails
any push that opens a non-localhost connection during dictation.
git submodule update --init --recursive
brew install cmake libomp rust
git config core.hooksPath Scripts/githooks
./run.sh # build + launch (./run.sh build to build only)See AGENTS.md for the full development guide and decisions.md for why
things are the way they are.
MIT — see LICENSE and ACKNOWLEDGMENTS.md for the excellent open-source work Rhino builds on.