yaowu-ai/snack-record

Snack Record is a macOS app that records meetings and turns them into Chinese language text transcripts, with everything happening on your own computer instead of being sent to a cloud service. It captures both the sound coming out of your speakers, such as other participants talking in a video call, and your own microphone, then uses a speech recognition model called FunASR to write out what was said, including punctuation and labels showing which speaker said what. You can also feed it existing audio or video files to transcribe rather than only live recordings. Finished transcripts are saved as plain text files to your desktop. The app offers a faster mode that skips speaker labeling and a more thorough mode that includes it, and for long recordings it shows a progress percentage and estimated time remaining. It can optionally watch for likely meetings starting in apps like Zoom, Tencent Meeting, WeCom, and Feishu, and pop up a local reminder to start recording, though this check happens only on your device. Installation is meant to be done with the help of an AI coding assistant that clones the repository and runs the included install script, which builds the app for your specific Mac and sets up a local only security identity so macOS trusts it, without needing an Apple Developer account. The first install downloads roughly two gigabytes of speech recognition models. The project states clearly that audio, transcripts, and task history stay on your machine and are never uploaded, and that screen video itself is never saved even though a screen capture permission is used to grab system audio. The app's own source code is released under the MIT license, but the speech models it downloads come from a separate service called ModelScope and may carry their own licensing terms.