Sokuji

Sokuji is a free, open source speech translator for people joining meetings across languages. It sends your translated speech through a virtual microphone, so other participants hear ordinary call audio and don't need to install anything. Their replies appear as translated subtitles on your screen.

The desktop app runs on Windows, macOS and Linux, and works with any app that accepts a microphone. It can also subtitle system audio, including videos, streams and recordings. The Chrome and Edge extension handles voice and captions within supported meeting sites, including Zoom, Google Meet, Microsoft Teams and Discord.

Local AI handles speech recognition, translation and speech output entirely on your device. After downloading models, you can use it offline without an account or API key, and your audio stays on your machine. It can use existing CPUs and integrated graphics. Local model choices include Whisper for recognition, Opus-MT and Qwen for translation, and Piper for speech.

For calls, Sokuji offers voice output with incoming subtitles or text-only translation in both directions. Provider-dependent voice cloning can preserve your voice, while adaptive speech speed helps translated speech keep pace. Movable subtitle overlays and conversation exports support presentations and meeting notes.

Cloud options include OpenAI, Google Gemini and Soniox, plus OpenAI-compatible endpoints. In cloud mode, audio goes directly to the chosen provider. The optional paid Kizuna AI account service manages credentials and billing; the app itself uses the AGPL-3.0 license.