LiveNotes 1.0.0 (Preview)

LiveNotes 1.0.0

LiveNotes records locally, displays English transcription and Chinese translation, and exports the recording with the current transcript.

Changes

• Preserve earlier sentences when live recognition revises a phrase, including shorter corrections and fragments within the same displayed second.
• Keep corrected final text from being overwritten or duplicated by an older speech draft.
• Finish recording by saving the live transcript, its trailing words, and translations already available. Normal save no longer starts another full-recording transcription pass or waits for missing translations.
• Preserve an existing Chinese preview when the final English sentence gains trailing words, labeling that saved preview as a partial translation.
• Export the current Markdown transcript and original audio in the background. Incomplete translation does not block export, and a failed replacement preserves the previous exported files.
• Start speech recognition before microphone capture and retain buffered text received around pause and stop.
• Remember the first-use recording acknowledgement. Later recordings start directly, including after restarting the app.
• Preserve preferences during regular Homebrew uninstall and updates. The Homebrew cask no longer runs privileged cleanup commands for obsolete runtime directories.

Recording, speech recognition, and translation continue to use Apple-native APIs. Speech and translation accuracy still depend on Apple's models. See the microphone and transcript validation report for measured results and limitations.

Install or update

brew install yongyaoduan/livenotes/livenotes

For an existing installation:

brew update
brew upgrade --cask yongyaoduan/livenotes/livenotes

Requires an Apple Silicon Mac with macOS 26 or newer. Low-latency translation requires macOS 26.4 or newer and installed translation assets. Existing saved recordings and transcripts remain compatible.

This build uses the project's existing preview distribution channel and is not Developer ID signed or notarized. If macOS blocks launch, open System Settings > Privacy & Security and choose Open Anyway.

Microphone and speech-recognition permissions are checked against macOS and requested only when their status is undetermined. An unchanged app normally reuses existing permissions after restart. Preview builds use ad-hoc signatures, so an updated binary may require macOS permission again; this release cannot guarantee permission reuse across updates. Stable Developer ID signing is needed for that. See the permission review.

Validation: 103 core tests passed. Local native audio and all release-required UI regression checks passed, including consent persistence and partial translation export. The hosted macOS native-recognition failure and complete test history are documented in the linked validation report. This package was built and validated on the local Mac.