Qwen Scribe offers private, local transcription and system-wide dictation for Apple Silicon Macs. It utilizes Qwen3-ASR running via mlx-qwen3-asr on the Mac's Metal GPU, ensuring all audio and transcript text remain on the local machine without requiring accounts or API keys.

Key features include drag-and-drop audio/video transcription in a local web interface, automatic language detection, word timestamps, SRT export, and local transcript history. For dictation, users can hold the right Command key in any text field to speak, and the transcribed text will be pasted at the cursor. A non-focus-stealing HUD provides status updates.

Requirements include an Apple Silicon Mac running macOS 14 or newer, Python 3.12+, and ffmpeg. The 1.7B model requires approximately 3.4 GB of unified memory. Currently in beta (v0.1.0-beta.1), core transcription and dictation are functional. Signed and notarized binaries are planned for a future release.