Muesli 0.8.3

Muesli 0.8.3

Muesli 0.8.3 brings dictations and meetings into one local timeline, adds Apple's newest on-device speech technology on supported Macs, and makes meeting context easier to keep with the conversation.

One timeline for dictations and meetings

• Timeline combines recent dictations and meetings in chronological order, with filters for source, destination app, and date.
• Dictations can show the app that ultimately received the text, making it easier to find work from Mail, Notes, browsers, and other apps.
• Existing Dictations and Meetings views remain available when you want a focused history.

Apple Speech on macOS 26

• Macs running macOS 26 can use Apple's system-managed SpeechAnalyzer for dictation, meeting transcription, and imported recordings.
• Apple Speech runs on-device, requires no separate Muesli model download, and follows the existing cleanup, diarization, summary, and persistence pipelines.
• Keep the current macOS language automatically or select another language supported by Apple Speech.
• On unsupported systems, Muesli continues to use its existing local model defaults.

Keep people with the meeting

• Calendar organizers and attendees can be captured from EventKit and shown with the meeting.
• Add someone from Apple Contacts or create a minimal new contact directly from the meeting header.
• Search meetings by participant name or email address.
• Participant data remains local to this Mac. It is not included in CloudKit sync, exports, CLI output, or telemetry.

Faster, more resilient model setup

• Model downloads now resume interrupted transfers, validate downloaded files, check disk space, and show clearer download and preparation stages.
• WhisperKit and supported local backends expose explicit language choices where available.
• Qwen ASR and LocalVQE packaging paths have been hardened so installed models and acoustic cleanup load more reliably.

A more capable local CLI

• The bundled muesli-cli can transcribe supported audio and video files with multiple local backends and return agent-friendly output.
• Portable dictionary import and export make custom vocabulary easier to move between installations and automation workflows.

Meeting and dictation reliability

• Production iCloud text sync is restored for signed macOS builds, with safe reconciliation for affected 0.8.2 installations when the private iCloud account matches.
• Dictation becomes available after a stopped meeting finishes audio transcription and cleanup, while title generation or summarization continues in the background.
• Meeting microphone capture can recover from degraded or changing input routes, and system-audio capture now detects stalled taps more reliably.
• Dictation drains the final audio buffer before transcription to reduce missing last words.
• Dictation once again restores your previous clipboard shortly after pasting, while preserving anything newer you copy and keeping app attribution outside the paste-critical path.
• iCloud text sync now uses Apple's durable sync engine while preserving Muesli's local outbox, improving recovery from interrupted syncs, account changes, and large pending histories.
• The floating recording indicator has more predictable drag behavior across displays and appearance settings.

macOS polish

• Added standard macOS window and navigation shortcuts, including direct access to Dictations and Meetings.
• Meeting notes render inline Markdown more naturally while remaining editable.
• Invalid recording durations no longer distort words-per-minute statistics.

Privacy and compatibility

• Audio, transcripts, meeting titles, contacts, file paths, hardware identifiers, and raw errors are not added to telemetry.
• Muesli 0.8.3 requires Apple silicon and macOS 14.2 or later. Apple Speech requires macOS 26 and a system-supported SpeechAnalyzer language.