yuxino/Mimi

把电脑或麦克风声音转成实时翻译字幕。Live translated subtitles from computer or microphone audio.

Mimi – Real‑time subtitles & translation

Mimi is a cross‑platform desktop (and Android) application that captures audio from your computer, microphone, or both, sends it to a speech‑to‑text service, optionally translates the result, and shows the text as a floating subtitle overlay. It is aimed at watching movies, live streams, online courses, or any media where you want live bilingual captions.

What it does

Live caption overlay – a floating window that can display the original language, the translation, or both. The window can be moved, resized, recolored and set to “immersive mode” where only the text is visible and mouse clicks pass through to the underlying app.
Audio source selection – capture system audio, microphone input, or both. On macOS/Windows you can target a specific application; on Linux you can pick an individual playback stream.
Save output – subtitles can be exported as plain‑text files and the captured audio can be saved as WAV files (both need to be enabled manually).
Multi‑language support – choose the language for speech recognition and a different target language for translation.

How it works

Configure a provider – In Settings → Model & Service you add a speech‑to‑text or translation service (e.g., OpenAI Whisper, Google Speech, Apple Speech, Gemini, DeepSeek, etc.) and supply any required API keys.
Select audio & languages – Choose the audio source and the source/target languages in the Sound & Language panel.
Start subtitles – Turn on Real‑time subtitles on the Subtitles page, grant the necessary OS permissions, and the app begins streaming audio to the chosen cloud or local service and displaying the returned text.

Platforms & Installation

Desktop: macOS 13+ (Apple Silicon & Intel), Windows x64, Linux x86_64. Download the latest installer from the GitHub releases page.
Android: a separate app is provided; see android/README.zh-CN.md for installation and usage.

Key technical details

Built with Tauri (Rust + web UI) for a lightweight native wrapper.
Uses cloud or local speech services; the app itself does not contain a model, it merely forwards audio and receives transcriptions.
Supports Wayland on Linux and includes fixes for floating‑window synchronization.
Open‑source under the MIT license.

Getting help & contributing

The repository includes a detailed usage guide, troubleshooting section, and a contribution guide.
Issues can be reported on GitHub; the project lists several active contributors and acknowledges community testing.

Who might find it useful?

Learners watching foreign‑language videos who want instant subtitles.
Streamers who want to provide live captions for accessibility.
Anyone needing a quick way to capture spoken content and keep a text record.

All information above is taken directly from the project's README.