Overview

Sokuji is a real‑time translation tool that acts as a virtual microphone. It lets you speak in your native language and have your words translated into the language of the other participant or meeting platform. The recipient hears the translation through the meeting app, while the speaker gets subtitles or spoken output back in their own language.

The desktop application registers itself as an ordinary microphone in the operating system. This means any software that can select a microphone, Zoom, Google Meet, Microsoft Teams, Slack, Discord, Jitsi Meet, Whereby, Gather, Yandex, Telemost, or even games and streaming programs, can use Sokuji without extra configuration. The app also monitors the system audio, so video or recorded content can be subtitled in real time. The browser extension works only on supported web meeting sites, offering the same virtual microphone functionality within the browser tab. No audio routing or external hardware is needed; the tool operates entirely on the user's machine.

Sokuji supports two‑way conversation. When you speak, the tool translates your voice to the target language and plays it back to the meeting participants. Incoming speech from the other side is translated into your language and presented as subtitles or read aloud. Users can choose to receive only subtitles, only whispered speech, or both. Multiple translation modes are available: full voice translation, subtitle‑only, or a mix where you hear the translation and see the original text. The tool adapts the speech rate to keep the translated audio in sync, and it uses semantic turn detection to avoid cutting off speech mid‑sentence.

The application is free and open source. It runs fully on device without a subscription, and it works offline once the GPU‑optimized inference engine is downloaded. The tool can also record a short clip to preserve the user's natural voice in the translation, choosing from a library of synthesized voices provided by the platform.