Type with your voice, on your own terms.

Why "Cadenza"?

Say it as it comes. Write it where you are.

In music, a cadenza is the passage where the soloist plays alone and freely. Speaking to your computer should feel like that: you say it your way, and the app follows.

The Chinese name 随言 carries the same idea as the tagline 随口说，随处写.

Typing is the slow part

You talk, and the words appear at your cursor. All the work happens locally, and anything that would leave your computer needs your explicit yes.

It is for chat, notes, email, search, code comments: the many small moments a day when typing slows you down.

What happens when you speak

Four steps, all local. The cloud is an option you choose, not a requirement.

1. Hold the shortcutLeft Option by default.
2. CaptureAudio stays in memory and is never saved.
3. RecognizeA local model, or the provider you chose.
4. WriteAt the cursor, or kept for you to copy.

Three principles

1. Local first, cloud by consentNothing leaves the Mac unless you turn it on. Cloud services get audio only after you agree for that provider, and the privacy text says what cancelling can and cannot undo.
2. Honest about limitsWe separate what was tested on a real Mac from what was only tested with fakes, and publish how accuracy is measured. Compatibility with every app is not claimed.
3. Built togetherMIT licensed, with a contributing guide, a code of conduct and a private way to report security problems. People review every change, and the privacy contract applies to all of them.

Platforms

Cadenza is built on macOS-only interfaces. Other systems would be separate projects.

Why only macOS, and what a port could reuse

License and credits

Cadenza is free and open source under the MIT license. Source on GitHub

Built on the work of others

Third-party notices and licenses