Every word of the meeting, written down as it's said. Nothing leaves your Mac.

What's said becomes text, and the text becomes a summary — or two with Pro, each in its own mode and language.

The window has up to three columns. On the left, completed transcript entries appear with timestamps. Text still being transcribed appears in a lighter bubble. The middle and right columns show two summary streams that update every few lines, each in your chosen mode: a short summary, a conversational summary, keywords, or a translation — and each in its own output language. Follow a short English summary in one while a Japanese translation runs in the other.

You can record from your microphone, from system audio, or from both at once, so a Zoom or Google Meet call comes through with the other participants and your own voice.

By default, transcription and summarization don't use the network.

Transcription and summarization run on Apple's speech recognition and Apple Intelligence, or on WhisperKit and MLX models running locally on the Mac. That makes it usable where cloud transcription isn't:confidential negotiations, interviews, clinical or legal consultations. It works on a plane. Pro adds the option of cloud language models (OpenAI, Google, Anthropic) with your own API key — used only when you explicitly switch to one, never by default.

Otonota looks at the chip and memory in your Mac and picks a sensible pair of engines to start with. An 8 GB MacBook Air runs on Apple's built-in engines; 16 GB and up unlocks larger on-device models. You can change engines at any time.

Drag an audio or video file onto the transcript column and it goes through the same transcription and summarization as a live recording. Pro

A voice memo, a Zoom cloud recording, a video of a lecture. Drop the file, or pick it with "Choose file", and m4a, mp3, wav, mov, mp4, and similar formats run through the same engines and the same two summary streams as a live recording. Long files show their progress, and you can start reading before they finish. An imported session can later be re-transcribed with a different engine, and the original is kept.

Stream 1 and Stream 2 can each be popped out into their own window, or shown side by side in one. Pro

Send a running summary of what's being said, and a translation into another language, to the meeting-room display or the lecture-hall projector. Participants can follow along without looking up from their own notes, and colleagues who don't speak the language can read the translated side. The transcript itself stays off the screen — only the summary and translation go up — so a slip of the tongue or a misheard word is never blown up in front of the room.

Each window has its own font, text size, and target display, and full screen hides every button and menu. The settings are saved per window, so the next meeting opens the same way. The auto-generation notice stays up for the first three minutes and then reappears briefly at the interval you choose — a way to tell the audience "this is generated automatically" without being a nuisance.

Follow a meeting held in Japanese through an English summary, or keep English notes from a lecture given in German. The interface, summaries, and translations come in these 22 languages. You can add product names, people's names, and internal jargon to a glossary to improve both the transcript and the summary.

• Japanese
• English
• Chinese (Simplified)
• Korean
• Spanish
• French
• German
• Italian
• Portuguese
• Russian
• Arabic
• Hindi
• Vietnamese
• Thai
• Dutch
• Swedish
• Norwegian
• Danish
• Finnish
• Polish
• Turkish
• Indonesian

A red dot marks the 10 languages the Mac's built-in engines can also transcribe. The other languages are transcribed by WhisperKit, which is included in the free version. The interface, summaries, and translations cover all 22.

When a recording or import is done, export the audio and the transcript and summary text, and, with Pro, JSON, CSV, and SRT with elapsed times from the start of the recording — plain files you can keep, not a format you're locked into. And if the app crashes mid-meeting, the recording and transcript survive: they're saved to disk every few seconds, and Otonota offers to pick up where it left off.

Record audio, transcribe and summarize it with the Mac's built-in engines, copy the text, and export your audio — all free, with no time limit.

The Mac's built-in engines, plus one WhisperKit model and one MLX model, are available for free. The other engines require Pro. You can export your recording and the plain text of the transcript and summaries for free; exporting data with elapsed times requires Pro. Pro is a one-time in-app purchase, not a subscription. Buy it once and use it on your Macs signed in with the same Apple Account. For reference, Pro is US$49.99 in the US (before tax) and ¥7,800 in Japan (tax included). Apple sets the price in each local currency elsewhere, so check the app or the Mac App Store product page for the price in your region.

Engines and languages

The 10 languages the built-in transcription covers are English, Japanese, Chinese, Korean, German, French, Spanish, Italian, Portuguese, and Russian. If Apple Intelligence is unavailable on your Mac, Otonota offers to download MLX Qwen3.5 2B and use it for summaries. The larger MLX models are listed only if they fit in your Mac's memory. Cloud speech recognition and language models use your own API key; the providers' usage fees are not included in Otonota's price and are billed directly to your own account.

Does any audio get uploaded?

Not by default. With Pro you can enable cloud speech recognition with your own API key, which sends audio to the provider. If you enable a cloud language model instead, only text is sent — never audio. There is one exception: if you choose Apple Speech for transcription and on-device recognition for that language isn't available on your Mac, macOS may recognize the audio on Apple's servers as part of the operating system. If that matters to you, choose Apple Intelligence transcription or WhisperKit.

Will it capture the other side of a Zoom or Meet call?

Yes. Record from your microphone, from system audio, or from both at once (microphone + system audio), which puts the other participants and your own voice into one transcript.

Which Macs?

Apple silicon Macs on macOS 26 or later. 8 GB of memory is enough for the built-in engines. On first launch Otonota looks at your chip and memory and picks engines and settings that won't strain the machine; you can then adjust the model and the summary window (how many segments go into one summary) at the top of each pane to match your Mac. Pro's local models only appear in the list if they fit in memory, but fitting is not the same as running smoothly: larger models take longer to summarize. Pick one that suits your Mac.

Which languages?

The interface, summaries, and translations come in 22 languages. Apple Intelligence writes summaries in 15 of them; the small MLX model included in the free version covers all 22. The built-in engines transcribe 10 languages (English, Japanese, Chinese, Korean, German, French, Spanish, Italian, Portuguese, Russian); WhisperKit large-v3, included in the free version, covers all 22. Pro adds the larger WhisperKit models.

Is there a time limit on the free version?

No. The free version has no time or usage limits.

Can I get my recordings out of the free version?

Yes. Exporting your recorded audio files and the plain text of the transcript and summaries (no timestamps) is free. Exporting JSON / CSV / SRT with elapsed times from the start of the recording requires Pro.

Windows?

No plans. Otonota is built around on-device AI on Apple silicon.

Where are the settings and engine choices explained?

In Help, in the order the app presents them.

Otonota is free on the Mac App Store. There's no account to create — install it, press record, and you're using it. Pro is available as an in-app purchase from the app's settings.

macOS 26 or later · Apple silicon (M1 and up)