Offline Transcription: Your Data Stays Yours.

Cloud transcription means your audio travels through the internet, gets processed on someone else's computers, and stored according to their policies.

Offline transcription is the simpler arrangement: the model runs on your device, so the audio has nowhere to go and nothing to wait for. On a Mac, the built-in Parakeet model runs at about 74× real time: an hour of audio in under a minute. There's nothing to trust because there's nowhere for the data to go.

It also works everywhere: planes, remote areas, offices with restricted networks. Without an upload step there is nothing to wait for before the text appears.

One note before we start: this guide covers transcribing recorded audio — imported files, meetings, interviews, voice memos. If you want to dictate text straight into any app instead, see our voice to text offline dictation guide.

Think of it like the Photos app on your phone. It doesn't need to send your pictures to a server to let you look at them. Whisper Notes works the same way: it turns sound into text right where you are.

The built-in Parakeet model comes with the app (the default — Moondream's Parakeet Ultra, post-trained from NVIDIA Parakeet V3, for English and 24 other European languages). The other models are optional downloads inside the app: Whisper (the multilingual option, 100+ languages; Large-V3-Turbo on Mac, Small on iPhone) and SenseVoice (the fast option for Chinese, Japanese, Korean, and Cantonese), plus Qwen3-ASR 1.7B (Beta) on Mac.

1. Tap record (or use the lock screen widget)
2. Speak
3. Stop
4. Text appears

On iPhone, with the built-in Parakeet model, the text already appears while you speak, in 25 European languages.

You can also import existing audio files. Meeting recordings, interviews and lectures all work, in any of the supported formats. Processing happens on the device, so there is no upload time to add.

Lock screen recording. One tap on the Lock Screen widget opens Whisper Notes and starts recording. Useful for catching a thought before it goes.

Audio import. Import existing recordings. A three-hour recording is one file, not 180 metered minutes; at about 74× real time with Parakeet on a Mac it finishes in under three minutes. No per-minute charge, no upload. You can drop in several files at once, and a video can come out as SRT subtitles.

100+ languages. Auto-detects what you're speaking.

Custom vocabulary. Add names and technical terms the model might miss.

Timestamps. Click any paragraph to jump to that moment in the audio.

Offline transcription fits if you:

• Want recordings to stay private
• Work in places without reliable internet
• Prefer one-time payment over subscriptions
• Transcribe sensitive conversations
• Want something that works without account setup

Offline processing has costs, and they are worth knowing before you commit:

Storage. The app is ~680MB. On Mac, the Whisper Large V3 Turbo model adds ~2 GB for highest accuracy.

First launch. Takes 3-8 seconds to load the model. That happens once per session, and every transcription after it starts immediately.

Platform. iPhone and Mac only.

Speed. Faster than real-time. On iPhone, the App Store listing reports about 18 seconds for five minutes of audio with Parakeet on an iPhone 15, and about a minute with Whisper. An M-series Mac is much faster: there, Parakeet runs at about 74× real time.

These trade-offs are worth it for some people. For others, cloud transcription makes more sense. Both have their place, and our guide to the best transcription apps for Mac and iPhone sets the offline and cloud apps side by side.

Price: $7.99 on iPhone. Free for 5,000 words a week on Mac, then $14. No subscriptions.

Models: Parakeet (default, built in), Whisper (Large-V3-Turbo on Mac, Small on iPhone) and SenseVoice on Mac and iPhone; Apple Speech on iOS 26 and macOS 26; on Mac also Qwen3-ASR 1.7B (Beta). All run locally on Apple silicon.

Privacy: No network requests during transcription. Verify with a network monitor.

Updates: Free.