Closed captions for your whole OS, translated on the fly

What it does mimi captures the audio playing through your machine — a film, a stream, a lecture — sends it to a speech-to-text provider you configure, and renders the result as a floating subtitle window showing the original text, a translation, or both. It runs on macOS 13+ (Apple silicon and Intel), Windows, and Linux x86_64, with a separate Android app, all built in Rust on Tauri.

The interesting bit The capture target is system audio, not your microphone, so it subtitles the video itself rather than whatever’s happening in your room — the difference between captions and a very confused transcript. There’s also no bundled speech engine: you bring your own provider credentials, which keeps the app vendor-neutral but means the heavy lifting happens in someone else’s cloud.

Key highlights

• Original text, translation, or both, with adjustable subtitle position, size, and color
• Click-through overlay window, so the subtitles never steal your clicks
• Optional local saving of subtitles and audio, exportable as TXT/WAV — off by default
• Cross-platform desktop app plus an Android port, MIT-licensed

Caveats

• Bring your own provider credentials; audio is sent to your configured speech service, so cloud charges and privacy are between you and that provider
• The README doesn’t name which speech providers are actually supported — you find out in Settings
• macOS requires Screen & System Audio Recording permission

Verdict If you consume content without caption tracks — foreign films, live streams, online courses — this is a genuinely useful tool. If you wanted a fully offline, zero-setup captioner, keep looking; the recognition lives in whatever cloud you bring.