Whisper Notes for Mac

The core feature: system-wide dictation. Hold the Fn key in any app—Gmail, Slack, VS Code, your terminal, Notion, Obsidian—and start talking. When you release, your speech becomes text.

No switching apps. No copy-pasting. Just hold Fn and speak wherever your cursor is.

Everything runs locally using Whisper Large-v3 Turbo, one of the most capable on-device speech recognition models available. Your audio never leaves your Mac. Setup takes about 30 seconds: Settings → Keyboard Shortcuts → Enable Global Dictation → grant Accessibility permission.

After recording finishes or importing audio, results stream out immediately—you don't have to wait for the entire transcription to complete.

This makes a real difference for longer recordings. Instead of staring at a progress bar, text appears paragraph by paragraph, so you see results faster.

Already have recordings? Drag and drop them from Finder. The app handles MP3, WAV, M4A, and most common formats.

This is useful for meeting recordings, lectures, podcasts, or interviews you've already captured. Everything processes locally—no cloud uploads, no waiting for servers. A MacBook M2 processes 10 minutes of audio in about 63 seconds. M4 chips are even faster, roughly 12x real-time speed.

Every transcript includes clickable timestamps. This is useful for long recordings.

Each paragraph gets a timestamp. Click any timestamp to jump straight to that part of the audio. Perfect for reviewing hour-long meetings or finding specific moments in lectures. You can export transcripts with timestamps included.

Add technical terms, names, or abbreviations that Whisper often misrecognizes. The model uses your custom words when transcribing.

Even obscure terminology, names, or jargon can be accurately recognized by the AI. It also helps preserve formatting—for example, ensuring "Claude Opus 4.5" instead of "claude opus 4.5", or "Whisper Notes" instead of "whisper notes".

Long silences used to cause the model to hallucinate—repeating phrases or making up text. VAD detects silence and handles it properly, keeping transcriptions clean.

With VAD, we've eliminated roughly 70% of the repetition issues caused by model hallucination in audio containing silence, significantly improving transcription stability and accuracy.

Cloud-based dictation sends your voice to remote servers. Every word you speak—personal notes, confidential meetings, private thoughts—travels over the internet.

Whisper Notes processes everything on your Mac. No network requests, no cloud storage, no data retention policies. If you're a journalist protecting sources, a lawyer with sensitive cases, a doctor discussing patients, or simply someone who values privacy—this matters.

It also works everywhere. On flights. In areas with poor connectivity. In secure environments that restrict network access.

MacWhisper is solid—Whisper-based transcription that works offline. Whisper Notes is faster, more accurate, and just $4.99 once for both Mac and iOS.

Wispr Flow is polished with cloud APIs that offer slightly better accuracy in some cases. The tradeoff: your audio goes to their servers, requires internet, and costs $120-180 per year.

Apple Dictation is convenient and free, but requires internet, has limited accuracy for technical content, and sends audio to Apple's servers.

The app runs in the menu bar, stays out of your way, and uses minimal resources when idle.