sunososobro-hub/tinglongwu

What it does

tinglongWu runs speech translation on your own machine: it listens to audio from videos, calls or meetings, transcribes it, translates Chinese/English/Japanese/Korean/Russian, and shows subtitles. In two-way mode it labels lines as me / them and can speak the translation into the call through a virtual microphone. Audio and text stay local, with no account or recording retention.

Install and first run

Windows 10/11 users download the setup executable and follow the installer. The setup fetches the speech and translation models once, so internet is needed only for that step. macOS and Linux (x86_64) use a curl | bash installer — macOS or Linux — and should read the script before running it.

After setup, choose Listen mode and play a meeting or video; subtitles appear on screen. For speak mode, route the virtual microphone into your call app, then speak in your language.

Requirements and limits

• Runs on a normal CPU; an NVIDIA GPU speeds it up substantially.
• Supported languages are Chinese, English, Japanese, Korean and Russian, with more planned.
• The full docs cover requirements, privacy and FAQ: documentation. Platform notes are in support.md.
• Licensed under MIT; copyright Alonso, 2026.

HotRepo's take

The project entered HotRepo's new-project list on 2026-11-14 with 11 stars gained in a roughly 1.5-hour window, reaching 112 stars and 23 forks. That is early traction, not evidence of long-term adoption. The local-processing pitch is clear, but the Windows installer and shell-pipe installers are the main distribution path and the README does not document how the binaries relate to the source, so verify the release artifacts before running them.