Version 2.0.1

MecoScribe 2 is a native Mac app. Drop in an audio or video file and get a speaker-labeled transcript you can play back, correct and reassign. The command-line tool is still included.

Changelog:

• New native SwiftUI Mac app with the same transcription pipeline as the CLI, including progress for each step and model downloads
• Editor with everything the HTML viewer offers: edit words and whole lines, insert and delete words, reassign speakers by selecting text (Edit Mode), drag words to a new position (Move Mode), rename, merge, add and delete speakers, undo/redo and Discard Edits
• Word highlighting synced to playback, follow playback, and a Jump Back button when the playing line is scrolled out of view
• Video sources play in a resizable pane above the transcript
• Projects are saved as a single .mscribe file (a ZIP with the transcript, word timings and the HTML viewer, schema 1.0); CLI output is imported on open
• Home screen with recent transcripts, Close Project, and a new app icon
• Update checker: the app checks GitHub Releases once a day and offers new versions (Settings › Updates, or MecoScribe › Check for Updates…)

Install:
Download MecoScribe-2.0.1.zip, unzip it and move MecoScribe.app to Applications. Requires macOS 14 or later on Apple Silicon. The app is not notarized, so the first time, right-click it and choose Open (or allow it in System Settings › Privacy & Security). Models download from Hugging Face on the first transcription.

Version 1.0.1

Changelog:

• Add support for Moondream Parakeet Ultra
• Add support for Nvidia Nemotron 3 Diarization