Mimi: Live System-Audio Subtitles and Translation for macOS and Windows

Watching a foreign-language video, joining an international meeting, or following a livestream without captions usually means switching between apps, searching for transcripts, or relying on subtitles that do not exist. Mimi takes a different approach: it listens to the audio output of your computer and turns that audio into live subtitles, with optional translation from a provider of your choice.

The result is a desktop overlay that can work with practically anything playing on the machine—not only a specific browser, media player, or video platform.

What Mimi does

Mimi is an open-source desktop application for Apple silicon macOS 13 and newer and Windows x64. It captures system output audio, sends that audio to the active translation or speech service, and displays the resulting transcription as an on-screen subtitle overlay.

Depending on the configured provider, Mimi can support different source languages, target languages, and quality modes. Its built-in description highlights translation into:

• Simplified Chinese
• English
• Japanese

Because the application operates at the system-audio level, it can be useful for videos, games, online courses, meetings, podcasts, and other media that may not provide captions themselves.

A flexible subtitle overlay

The overlay is designed to stay useful without taking over your screen. You can:

• Move the subtitle window to any position
• Resize it for readability
• Collapse it when you need more screen space
• Pause subtitle generation
• Enable click-through behavior so it does not intercept mouse input
• Use Immersive Mode for a more focused fullscreen subtitle experience

This makes Mimi suitable for both casual viewing and more practical workflows, such as keeping translated speech visible while working in another application.

Provider configuration without repeated setup

Mimi uses a bring-your-own-provider model. You supply the API credentials for the translation or speech service you want to use, and usage charges may apply according to that provider's pricing.

The application supports saved service configurations, allowing you to store and switch between providers without repeatedly entering credentials. Credentials are stored in the operating system's credential store rather than being handled through a Mimi account.

To configure Mimi:

1. Open Translation Service.
2. Choose a supported provider.
3. Enter the required credentials and service settings.
4. Save the configuration.
5. Switch providers later from the same settings area when needed.

This approach is particularly useful for developers and power users who already have API access, want to compare providers, or need separate configurations for different projects.

Privacy model

Mimi's privacy claims are centered on minimizing what the application itself retains and requests:

• No Mimi account is required.
• It does not use a microphone for capture.
• It does not capture the screen.
• Audio is not saved locally.
• Subtitles are not saved locally.
• Audio is sent only to the currently active provider.

The important qualification is that the active provider still receives the audio necessary to perform transcription or translation. Users should therefore review the provider's retention and privacy policies, especially when processing meetings, proprietary material, or other sensitive content.

Getting started

Download the latest release from the project's GitHub releases page. The repository provides:

• An Apple silicon DMG for macOS
• Windows x64 EXE and MSI installers
• Source code for building the project yourself

After installation:

1. Open Mimi and configure a translation service.
2. Start playing audio on your computer.
3. Open the Mimi menu bar icon on macOS or the system tray icon on Windows.
4. Select Start.
5. On the first macOS run, approve the requested Screen & System Audio Recording permission.

Although the application does not capture the screen, macOS requires this permission for system-audio recording. Windows users may see SmartScreen warnings because the preview EXE and MSI installers are unsigned.

Platform-specific notes

Apple silicon macOS

Mimi supports macOS 13 and later on Apple silicon. Releases provide an ad-hoc-signed DMG without Apple notarization. If macOS blocks the first launch, use System Settings → Privacy & Security → Open Anyway.

After updates, macOS may ask for recording or Keychain permissions again. This is expected behavior when the application identity or packaged build changes.

Windows x64

Windows x64 builds are available as EXE and MSI installers. Since the preview installers are unsigned, Microsoft SmartScreen may display a warning before installation.

In-app updates

Mimi includes an update workflow in its settings. Go to:

Settings → General → Software Update

The application downloads updates with visible progress and then provides an installation action. On Windows, Mimi reopens after the installation completes. On macOS, the update flow offers a separate Restart and Finish Update action.

One compatibility detail matters for existing users: versions older than v1.3.8 require one manual installation before in-app updates can be enabled.

The latest listed release is Mimi v1.3.12, published September 12, 2026.

Building and contributing

Mimi is built primarily with Rust and TypeScript, with the repository language breakdown listed as:

• Rust: 72.9%
• TypeScript: 18.5%
• CSS: 3.3%
• PowerShell: 1.9%
• Shell: 1.7%
• JavaScript: 1.7%

The project includes a Rust-based src-tauri application layer, frontend code under src, build scripts, contribution guidelines, and a security policy. Developers interested in modifying the overlay, adding provider integrations, or improving desktop lifecycle behavior should start with CONTRIBUTING.md and the project documentation.

Mimi is released under the MIT License, making it accessible for inspection, modification, and reuse within the license's terms.

Who should try Mimi?

Mimi is a strong fit for anyone who needs captions outside a platform's built-in ecosystem: language learners, international teams, accessibility-focused users, viewers of foreign media, and developers experimenting with speech and translation providers.

Its most compelling design choice is the combination of system-level capture and provider flexibility. Instead of integrating separately with every application that produces audio, Mimi creates one desktop-wide subtitle layer. Pair that with configurable overlays, saved service profiles, in-app updates, and a no-account privacy model, and it becomes a practical tool for turning almost any computer audio into readable, optionally translated text.

Source

yuxino/mimi: Live translated subtitles for anything playing on your Mac or PC.