AudioScript open-sourced: local real-time bilingual subtitle tool

On November 2, 2026, a developer posted on V2EX about AudioScript, an open-source local real-time bilingual subtitle tool. AudioScript listens to whatever a computer is playing and turns foreign-language videos, podcasts and meetings into floating bilingual subtitles with the original line on top and the translation below. Both recognition and translation run on the machine: no audio or text is uploaded, and no account registration is required.

In the post the author explains the motivation. When watching foreign-language videos or joining English meetings, subtitles are often needed, but existing options either push everything through the cloud, raising privacy, network and latency concerns, or force users to install a virtual sound card such as BlackHole to route system audio, which is fiddly to configure and can disturb the existing audio chain. AudioScript is meant to be open-and-use and to leave the system's audio output untouched.

For capture it reads system audio directly, via Core Audio Process Tap on macOS and WASAPI loopback on Windows, so no virtual sound card is involved. Users can also switch to the microphone or mix both sources. Recognition is streaming: partial results refresh in real time and each sentence is finalised at its end. A Chinese-English model ships built in, so the tool works offline right after installation. The post links the official site at https://audioscript.looplooptech.com, a video walkthrough at https://www.bilibili.com/video/BV1zwnY6kEAd and the GitHub repository at https://github.com/niyongsheng/AudioScript. As of now only this single post is available, so there is no later development to report.