Whisper Notes

An offline speech-to-text app for iOS and macOS. Record thoughts, meetings, and lectures, then transcribe them quickly and accurately—entirely on your device.

Yes—Whisper Notes runs the same open-source Whisper models (including Whisper Large V3 Turbo) 100% locally on your device, packaged as a native app. No Python, no whisper.cpp compilation, no command line required: models download inside the app, and everything runs offline on the Neural Engine.

Yes—completely. Recordings and transcripts are processed on your device, with no internet connection required.

No. Your recordings and transcripts never leave your device: transcription runs locally, we never see them, and they're never used to train AI.

Yes. Import recordings from meetings, lectures, or interviews. Supports most formats—MP3, M4A, WAV. On iPhone, you can also share recordings directly from the Voice Memos app: tap the recording → '...' → Share → 'Transcribe with Whisper Notes', and it keeps its original recording date. On iPhone you can also import several files at once, or share up to 20 from the Files app; they are transcribed one by one. To automate it, use the Transcribe Audio File action in the Shortcuts app: it takes an audio file and returns the text, with an Include Timestamps option. Other Whisper Notes actions there start, pause, resume and stop a recording, set a note's title, and get the last transcription (as text or as a file) or the last recording. Everything transcribes offline on your device.

Hold the Fn key in any app—Gmail, Slack, VS Code, Notes, ChatGPT, Claude, or any browser text field—and speak. Release the key, and your text appears. Works completely offline, so your voice stays private even when using cloud-based apps. To customize the shortcut key, open the left side panel → Voice Typing (not in Settings). This is useful if you're using an external keyboard without an Fn key.

Yes—on Mac. Whisper Notes detects when a Zoom, Microsoft Teams or Google Meet call starts and asks whether to record; you can turn on automatic start in Settings → Meeting Recording. Everything is transcribed locally on your device. Meeting recording needs Screen Recording permission (System Settings → Privacy & Security → Screen & System Audio Recording) — that's how the other participants are captured, from system audio, whatever you're listening on. AirPods and other Bluetooth headphones work: the far side is always recorded, and "Record microphone" (on by default) adds your own voice from your selected mic.

Yes. Whisper Notes supports on-device speaker identification—also called speaker diarization—for recordings, imported audio files, and meeting recordings. Rename a speaker once and the label updates throughout the transcript; SRT and VTT exports preserve speaker names. If voices sound similar, set the number of speakers yourself, up to 10. Available on iPhone and Mac.

100+ languages, including English, Chinese, Spanish, German, French, Japanese, Korean, Russian, Arabic, Hindi, Portuguese, Italian, Dutch, Turkish, Indonesian, Vietnamese, Thai, Czech, Polish, Ukrainian, Swedish, and many more.

iPhone: iPhone 12 and later (including all Pro/Plus/Max models and the third-generation iPhone SE), on iOS 18. iPhone 11, the second-generation SE and older models are not supported: their chips are too slow for on-device transcription.

Mac: All M-series Macs (M1, M2, M3, M4, and future chips). Intel Macs are not supported. We recommend 8GB+ RAM if you run the larger Whisper Large V3 Turbo model, the multilingual option covering 100+ languages.

No, and we are not building one. Whisper Notes runs on Apple's Neural Engine, so it works on iPhone and iPad (iPhone 12 or later) and on Apple Silicon Macs (M1–M4). There is no Windows, Android or Intel Mac version, and none is planned.

We distribute only through the App Store and whispernotes.app. If a site offers a Whisper Notes .exe, an APK or a "PC version", it is not ours — treat it as malware.

For offline transcription on Windows, Buzz is a free Whisper app with a normal window and buttons, and whisper.cpp is the command-line route. Both run the models locally, so your audio stays on your PC. More detail on our Windows page.

Whisper Notes offers multiple AI models optimized for different use cases.

On iPhone: Parakeet (default) supports 25 European languages with fewer errors. Whisper supports 100+ languages but may occasionally add filler words in quiet sections. On iOS 26 you can also select Apple Speech, the engine built into the system.

On Mac: Parakeet (default) is optimized for 25 European languages. Whisper Large V3 Turbo is the multilingual option, covering 100+ languages — pick it when your audio falls outside Parakeet's 25. On Mac, in both the Mac App Store version and the Mac app from our website, you can also select Qwen3-ASR 1.7B (Beta): 30 languages, text appearing as it transcribes, 16 GB of memory recommended — it is not on iPhone. It is the most accurate choice for Chinese, Japanese, Korean and Cantonese. SenseVoice is the fast option for those languages, and the one to use on iPhone; for Japanese and Korean, Whisper Large V3 Turbo is close to SenseVoice or better. On macOS 26 or later, Apple Speech is a fifth option. The app does not switch engines for you: you select one in Settings.

On both, the built-in Parakeet model is Moondream's Parakeet Ultra, post-trained from NVIDIA Parakeet V3.

If transcription quality is poor for your language, try switching models in Settings. Also try manually selecting your language instead of using auto-detect. To see which model makes the fewest mistakes in your language, check our per-language model comparison.

With Parakeet (the default model), transcription is very fast: on Mac it runs at about 74× real time, so an hour of audio takes under a minute. On iPhone, speed depends on your device: newer models like iPhone 15 are significantly faster than older ones. If you switch to Whisper for broader language support, expect roughly a 1:5 ratio—5 minutes of audio in about 1 minute on iPhone 15.

Partly. A recording is transcribed when you stop it, and an imported file when you import it; the text appears gradually as it processes. When you record a meeting on Mac, a live transcript panel shows the text during the meeting, updated about every 30 seconds. With voice typing in the Mac app from our website, hold Fn, speak, and the text is pasted when you release Fn.

Yes. You can export a transcript as text with timestamps, as SRT subtitles, or as Markdown. On Mac you can also export VTT; on iPhone, VTT export appears once speaker detection has run on the note. On Mac, an imported video opens in a subtitle workspace where you can preview the captions over the video, edit them and export SRT or VTT. The Mac also saves every transcript as a Markdown (.md) file next to its audio, with YAML front matter, in a folder you choose; an Obsidian vault works. On iPhone, turn on Auto-export Transcripts in Settings › Data Management and choose a folder (iCloud Drive, On My iPhone, Dropbox): each new transcript is saved there as a .txt file, and with Include Audio the recording goes next to it under the same name. It only exports; edits and deletions in the app never change those files.

Yes, in the Mac app from our website and on iPhone. On Mac, after you download the on-device AI model (Gemma 4), click Summarize to get a summary, ask for action items, or chat with your transcript to ask questions or extract key points. It can also generate titles.

On an iPhone with iOS 26 that supports Apple Intelligence, tap Summarize and Apple Intelligence writes the summary on the device. Either way, nothing is uploaded.

Yes—in the Mac app from our website. For recordings longer than 3 minutes, you can chat with a local AI about your transcript—review key points from meetings, extract action items, or ask questions about what was said. To access it, tap "View Detail Transcript" on any recording over 3 minutes. Everything runs on-device, so your conversations stay private.

For shorter recordings, Whisper Notes performs auto clean-up but the chat option is not available.

Yes — the Whisper Notes Mac app downloaded from our website includes a whispernotes command-line tool and a built-in MCP server, so Claude and other AI assistants can search your history, read transcripts, and transcribe files. Everything runs on-device over a local socket with no open ports. Full setup guide: Whisper Notes CLI & MCP

One-time purchase—no subscriptions, no ads, no hidden costs.

iOS: Buy directly from the App Store for $7.99.

Mac: Download free from our website. You get 5,000 words a week free—no credit card, no sign-up. When you're ready, unlock the full version for $14 inside the app. All future updates included on both platforms.

Get the iOS and Mac App Store versions from the App Store. Download the Mac app from our website, free for 5,000 words a week. We recommend it if you want system-wide Fn voice typing. Current prices are shown in each purchase channel.

On Mac, yes: the Mac app from our website is free for 5,000 words a week, every week, with no time limit. No credit card, no sign-up. Use it for meetings, lectures, voice memos—decide for yourself.

On iOS, the App Store handles purchases directly at $7.99. If you're not satisfied, request a refund through Settings → Request Refund.

Short answer: not for the App Store versions; yes for the Mac app from our website. One App Store purchase includes both iOS and the Mac App Store version, and both support recording and transcribing imported audio.

Because of Apple’s restrictions, the Mac App Store version can’t offer system-wide voice typing (hold Fn to speak). The Mac app from our website also adds the on-device AI assistant. If you need those capabilities, choose the separately sold Mac app from our website. Thank you for supporting independent development.

Yes. On iPhone, go to Settings → Request Refund. For the Mac app from our website, go to Settings → Feedback & Support → Request Refund, or use whispernotes.app/refund. The Mac App Store version is refunded by Apple: request it at reportaproblem.apple.com.

No. This is a deliberate privacy decision. Since all your recordings and transcripts stay on-device, there's no cloud sync between your iPhone and Mac. This means your audio never passes through any server—not ours, not Apple's. You can manually export transcripts if you need them on another device.

It depends on your device. Everything runs on your iPhone, not in the cloud, so the chip sets the speed: a newer iPhone finishes sooner. With Parakeet on an iPhone 15, five minutes of audio takes about 18 seconds; Whisper takes about a minute. The first transcription after an app update takes about 20 seconds longer while the model compiles once. Whisper Notes needs iPhone 12 or later.

iOS limitation. Whisper Notes needs GPU access to transcribe, and Apple doesn't let apps use the GPU in the background. If you switch apps mid-transcription, come back and tap the retry icon on the note, or choose Re-Transcribe from its ⋯ menu.

Here is what it gets used for, concretely. Interviews: record on iPhone, or drop an existing file into the Mac app, and the transcript stays on the device that made it. Meetings: on Mac it captures system audio and the microphone together, so nothing has to join the call as a participant. Drafting: in the Mac app you download from our website, hold Fn in any app — Mail, Notes, an editor, a terminal — speak, release, and the text lands at the cursor. Voice memos: one tap on the Lock Screen widget opens the app and starts a recording.

Because there is no account and nothing is uploaded, it also suits recordings that are not allowed to leave the machine under an NDA or a data-handling policy. One thing to check before you buy: system-wide Fn voice typing is only in the Mac app from our website; Apple's rules keep it out of the Mac App Store version.

It depends on what you need, so here is what is checkable. Transcription runs on your iPhone or Mac and the audio stays there, which is why the app works in airplane mode. It is a one-time purchase with no subscription: $7.99 on iPhone, and on Mac 5,000 words a week free, or $14 once for unlimited. You choose which engine runs. On Mac, both the Mac App Store version and the Mac app from our website have five: Parakeet, the default, covers 25 European languages; Whisper (Large-v3 Turbo or Small) covers 100+ languages; SenseVoice is the fast option for Chinese, Japanese, Korean and Cantonese; Qwen3-ASR 1.7B (Beta) covers 30 languages, needs 16 GB of memory (recommended), and is the most accurate of them for Chinese, Japanese, Korean and Cantonese; Apple Speech is the system engine on macOS 26 or later. iPhone has Parakeet, Whisper Small and SenseVoice, plus Apple Speech on iOS 26. The interface is available in 32 languages.

Whisper Notes fits when the recording has to stay on the device and you would rather pay once. A cloud tool fits better when you need a shared team workspace, collaborative editing, or a Windows client. Our comparison of the best offline transcription apps says which app wins in each of those cases.

Yes. Both offer Fn key dictation on Mac. Typeless uploads your voice to cloud servers and charges $12–$30 a month. Whisper Notes runs on your Mac for $14 once, and your voice never leaves it. In our test on an M4 Pro, a 20-second clip came back in 2.5 seconds, against 7–10 seconds for Typeless. Trade-off: Typeless rewrites your text with a cloud LLM. Whisper Notes punctuates as it transcribes, and in the Mac app from our website, with AI features turned on, the local Gemma model can remove filler words. It does not restructure your sentences, so the transcript stays close to the recording.

Yes. Both can run the same Whisper Large-v3 Turbo model on-device. MacWhisper Pro costs €64 one-time on macwhisper.com (Mac-only — the Mac App Store version is a separate app called Whisper Transcription). Whisper Notes is $7.99 once on iPhone; on Mac, the first 5,000 words every week are free, then $14 once. Its interface is simpler. Trade-off: MacWhisper offers 50+ export formats and video sync for professional audio workflows.