Questions,
answered straight.

What leaves your Mac, what does not, and how to use Mac Note Taker for sensitive meetings.

• Where does my audio go?

  Audio stays on your Mac. Mac Note Taker records to a local PCM WAV file, runs transcription on the Apple Neural Engine, and writes the result to a local SwiftData store. The raw audio file never leaves the disk unless you export it yourself.

• Does Mac Note Taker send anything to your servers?

  Only a license check, roughly once a day, which sends your license key and machine fingerprint. No audio, no transcript, no meeting titles, no calendar data. There is no analytics SDK and no telemetry.

• Can I use this for HR, legal, or NDA meetings?

  Yes, and that is the most common use case. Run the AI Assistant on Ollama instead of OpenAI and the entire pipeline stays on your Mac: capture, transcription, diarization, and summary. Nothing crosses the network.

• What about GDPR?

  Mac Note Taker is not a data processor for your meeting content because we never receive it. You are the controller and the data lives on the Mac you own. The only personal data we hold is the email used at checkout and the license key tied to it.

• Do you train models on my data?

  No. We have no servers that receive transcripts and no training pipeline. The app is a one-time license, not a data product.

• Where are recordings stored on disk?

  Under your home directory, inside the app sandbox container at ~/Library/Containers/com.macnotetaker.app/Data/Documents. You can move the library or back it up like any other folder.

How Mac Note Taker captures both sides of a call without a bot or a virtual audio driver.

• Does it record both sides of Zoom?

  Yes. Mac Note Taker uses Core Audio process tap to capture the system audio output of Zoom, Meet, Teams, Slack, Discord, and any other app, plus your microphone, in two separate channels. Both sides land in the same transcript with separate speakers.

• Do I need a meeting bot?

  No. There is no bot that joins the call. The recording happens on your Mac from your seat in the meeting, which is why it works for huddles, in-person rooms, and audio coming out of any app, not just video calls.

• Do I need a virtual cable, Loopback, or BlackHole?

  No. Mac Note Taker uses the official Core Audio process tap API on macOS 14.2 and newer, so no kernel extensions or audio routing tricks. You do not install BlackHole, Loopback, or SoundFlower.

• Does it work with Google Meet, Teams, Slack huddle, and Discord?

  Yes. Mac Note Taker taps system audio output, so any app that plays sound through macOS is captured. That covers Zoom, Google Meet, Microsoft Teams, Slack huddles, Discord, FaceTime, Webex, and browser-based calls.

• Does it work on Intel Macs?

  No. Mac Note Taker requires Apple Silicon (M1 and newer). The on-device speech and diarization models run on the Neural Engine, which Intel Macs do not have.

• What macOS version is required?

  macOS 14.2 or newer. The Core Audio process tap API used for system audio capture shipped in 14.2, which is why earlier versions are not supported.

• Can it auto-detect meetings?

  Yes. With Calendar access granted, Mac Note Taker can offer to start recording when a scheduled meeting begins, or detect that a meeting app is active and prompt you. It never starts a recording silently.

• What audio format is used?

  PCM WAV, mono, 16 kHz, separated by source. We do not use AAC or m4a because a corrupted m4a header can lose the entire recording, which has happened to real customers. WAV degrades gracefully.

Bring your own model. Mac Note Taker does not run inference on our servers.

• What AI models can I use?

  Any local Ollama model, or any OpenAI-compatible endpoint. That includes OpenAI, Groq, Together, Fireworks, Azure OpenAI, and self-hosted gateways like LiteLLM. You set the base URL and key in Settings.

• Do I need Ollama?

  Only if you want full local inference. If you are happy sending the transcript to your own OpenAI tenant, you do not need Ollama at all. Most people install Ollama later, once the muscle memory of having a summary every meeting kicks in.

• Can I use OpenAI?

  Yes. Paste your OpenAI API key into Settings and pick a model like gpt-4o-mini. The request goes from your Mac directly to api.openai.com. We do not proxy it and we cannot see the transcript.

• Does my transcript go to OpenAI?

  Only if you choose OpenAI as the provider. In that case the transcript and prompt go to your own OpenAI tenant under your key. If you pick Ollama, nothing leaves the Mac. The choice is per provider, set once in Settings.

• Is the AI required to use the app?

  No. The AI Assistant is optional. Recording, transcription, diarization, speaker rename, and search all work without any model configured. Summaries and action item extraction are the only features that need a model.

• What does the summary actually contain?

  A 3 to 5 bullet recap, action items grouped by owner, 3 to 5 topical sections, and speaker rename suggestions pulled from the transcript and calendar attendees. You can edit the prompt template in Settings.

One price, three Macs, no subscription, no account.

• Is Mac Note Taker a subscription?

  No. It is a one-time $149 lifetime license. You buy it once and own it. There is no monthly fee, no per-meeting fee, and no usage limit.

• How many Macs can I activate?

  Three. The license is tied to a machine fingerprint and you can activate up to three Macs at the same time. You can deactivate from Settings if you need to move the license to a new machine.

• What is the FOUNDER coupon?

  FOUNDER is a Stripe coupon that takes $70 off the lifetime price, dropping it from $149 to $79. It is capped at the first 100 redemptions and disappears once the cap is hit.

• Refund policy?

  14-day refund, no questions, minus a 5% processing fee that covers the Stripe cut we do not get back. Email support and we issue it the same day.

• Will the lifetime license get future updates?

  Yes. Lifetime means all future versions of Mac Note Taker on macOS. Updates ship through Sparkle, in-app, and are included as long as the app exists.

• Do I need an account to use the app?

  No. There is no account, no signup, no login. Your license key is the identity. Paste it into Settings and the app activates.

• Can I expense it or get a receipt?

  Yes. Stripe emails a tax-compliant receipt at checkout. Reply to support if you need a custom invoice with a company name or VAT ID and we will reissue.

Which ASR model runs, which languages it covers, and how speaker memory works across meetings.

• What ASR model is used?

  You pick from Parakeet TDT v3, Whisper Large v3, Distil-Large v3, and Whisper Small.en. Parakeet is the default because it is fastest on the Neural Engine and on par with Whisper Large for English meetings.

• Which languages are supported?

  Parakeet covers 25 European languages. The Whisper models cover 99 languages with varying quality. You can switch model per meeting if you record calls in multiple languages.

• Can it identify the same speaker across meetings?

  Yes. Mac Note Taker stores a voice fingerprint per speaker locally and matches new recordings against it. If Anya talks in Monday's standup and Friday's review, she shows up as Anya in both, not Speaker 2.

• Can I rename speakers?

  Yes. Right-click any speaker label in the transcript to rename. The rename applies retroactively to that meeting, propagates to future meetings via the voice fingerprint, and can also be done in bulk by the AI Assistant using calendar attendees.

• How accurate is the transcription?

  On a clean Zoom call with Parakeet, word error rate is typically 4 to 7 percent. On noisy in-person audio it can drop to 10 to 15 percent. Switch to Whisper Large for higher accuracy at the cost of slower processing.

• Can I edit the transcript?

  Yes. The transcript is editable inline. Edits are stored alongside the original so you can revert. Summaries re-run against the edited version.