I built an open-source voice input for macOS that never uploads your voice
I built an open-source voice input for macOS that never uploads your voice Most voice input tools upload every recording to somebody's server. I wanted something simpler: hold a hotkey, talk, release — and the text appears at my cursor. With my voice never leaving the Mac. So I built Cadenza (随言). It's MIT open source, macOS 14+, Apple Silicon and Intel. Repo:…
A software developer has created an open-source voice input tool for macOS that ensures user privacy by keeping voice recordings on the device itself. The tool, called Cadenza, operates without uploading any voice data to external servers, unlike many other voice input tools. It is compatible with macOS 14 and supports both Apple Silicon and Intel processors.
The open-source code is available on GitHub (https://github.com/DragonKingIO/Cadenza-voice) and can be found on the project's website (https://dragonkingio.github.io/cadenza-site/).
Cadenza's functionality is straightforward. Users hold a designated shortcut key, the app captures audio from the microphone, and text is generated based on the recognized speech. The text appears at the cursor once the shortcut is released. The app supports various voice recognition models, which can be downloaded within the application with checksum verification, ensuring the integrity of the models.
The tool does not require a network connection for inference, meaning no voice data leaves the user's Mac. Users have the option to use cloud providers for recognition if desired, with options such as OpenAI, Groq, Google, Azure, AssemblyAI, ElevenLabs, and any service with an OpenAI-compatible endpoint. The developer emphasizes that users retain control over their data, as they provide their own API keys if they choose to use cloud services.
Additional features of Cadenza include text cleanup (removing filler words), AI-based text polishing, AI speech translation, on-device voice translation, and screenshot OCR. While still in an early preview stage, the developer reports using Cadenza daily in various environments. However, the tool has not been notarized by Apple, requiring users to manually allow the app to run each time it is launched.
The developer seeks feedback on recognition accuracy across different accents, microphones, and real-world settings, as well as cloud provider integrations, UI translations, and documentation in other languages. The project welcomes support in the form of stars, as well as issues and pull requests for further improvement.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
