Pricing & licensing
Is Open Whisperer free?
Yes. Open Whisperer is free and open source under the MIT license — no subscriptions, no accounts, no API keys, and no word limits. You can read, build, and modify the full source on GitHub.
Privacy
Does my voice or data leave my Mac?
No. Speech-to-text (WhisperKit) and text-to-speech (Kokoro and Supertonic-3) all run locally on your Mac, in-process on the Apple Neural Engine. There are no cloud APIs, no telemetry of your audio, and nothing leaves your machine. It works fully offline.
What permissions does it need?
Microphone to record your voice, Accessibility to type transcribed text into other apps, and Speech Recognition for on-device hands-free keyword detection. All processing stays on the Mac.
Compatibility & requirements
Which AI tools does it work with?
Spoken replies work with Claude Code, Codex, Antigravity, and Pi. The Agents tab lists all four with their own Connect button, so you can connect as many as you use — Connect wires up a speak tool (a drop-in extension for Pi) so every response is read aloud. Dictation works with any macOS app: your voice types into whichever app is focused, and auto-focus can bring a chosen app forward first.
What do I need to run it?
An Apple Silicon Mac (M-series) running macOS 14 or later. It's fully native — no Python. On first launch it downloads the on-device Whisper and Kokoro models and runs them in-process on the Apple Neural Engine.
How many languages and voices does it support?
Speech-to-text covers all 100 languages Whisper supports, in a searchable picker grouped by how well the model actually does: good accuracy (56 languages at 35% word errors or fewer), limited accuracy (9, each badged with its own rate), and untested (35 with no published benchmark). Nothing is hidden — the rates come from OpenAI's published large-v3 benchmarks, so you're told what to expect rather than having the choice made for you. English is the default, since pinning a language beats auto-detect on the short phrases dictation produces; auto-detect is still there as an explicit choice.
Text-to-speech offers 102 voices across 32 languages, selectable in Settings without a restart: the full 54-voice Kokoro roster, plus every one of the 24 languages Kokoro has no voice for — Dutch, German, Greek, Polish, Russian, Ukrainian and the rest. Each row shows the language, the speaker and whether it still needs downloading, and a preview button reads a sample in the voice's own language. Pick one of those and your agent writes its spoken reply in that language.
Why does macOS say the app can't be opened?
The app isn't notarized (no paid Apple Developer signing), so Gatekeeper blocks it on first open. Clear the quarantine flag with xattr -cr /Applications/OpenWhisperer.app and it will launch normally.
Using it
How do I talk to it?
Three ways: Press-to-Talk, Hold-to-Talk, or fully hands-free. In hands-free mode, say "initiate" to start, stay silent for a few seconds to submit, and say "hold on" to interrupt while it's speaking.
Does it read AI responses aloud?
Yes — this is the part built-in dictation tools skip. Replies from Claude Code, Codex, Antigravity, and Pi are spoken with streaming text-to-speech that starts on the first sentence and can begin mid-turn through an in-app speak tool. A Response mode lets you choose when replies are spoken: dictated turns only (the default) or always.
How is it different from macOS Dictation?
macOS Dictation only types text into a focused field. Open Whisperer adds streaming text-to-speech that reads your AI's replies aloud, hands-free control with wake words, and one-click Claude Code, Codex, Antigravity, and Pi integration. See the full Open Whisperer vs. macOS Dictation comparison.