Open-Source Mimi Brings Live Subtitles to Every App on Your Desktop

Mimi is an open-source live-subtitle app that captures audio from the operating system, sends it to a speech-recognition service, translates the transcript, and displays the result over any application. That system-level approach covers streams, meetings, lectures, games, and local media without depending on site-specific browser extensions.

The MIT-licensed project uses Rust and Tauri and has attracted more than 550 GitHub stars. Desktop releases support macOS 13 or later on Apple silicon and Intel, plus Windows and Linux on x86_64. A community contributor maintains a separate Android build. Downloads and documentation are available on the project site.

One audio tap, every app

Mimi captures system audio, microphone input, or both, with system audio selected by default and microphone access requiring an explicit choice. On macOS and Windows 11, users can limit capture to one application instead of recording the complete system mix.

1. Capture: Read audio from the system, a microphone, or a selected application.
2. Recognition: Send the audio to the configured speech-to-text provider.
3. Translation: Pass the transcript to an independently configured translation service.
4. Display: Render the original transcript, the translation, or both in a floating overlay.

The overlay supports configurable position, text size, and color. A click-through mode lets mouse input reach the application beneath it, which is useful for games, presentations, and other full-screen software. Mimi can also save subtitles as TXT files and captured audio as WAV files; both recording options remain disabled until the user enables them.