Questions, answered

Yes. Tape is free to download and use, with every feature included. There is no account, no trial, no pro tier and no subscription. If it saves you a paid service, a coffee is appreciated and never required.

No. Recording, transcription, summarization and chat all run on your machine. Tape's only network calls are the update check on this site and the one-time model download. Neither one carries your audio, transcripts or summaries. The privacy page lists every request the app can make.

macOS 15 (Sequoia) or newer on Apple Silicon. Transcription and the AI use the Neural Engine and GPU; an M1 handles it comfortably. Plan for a few gigabytes of disk for the models, plus the recordings themselves at roughly 30 MB per hour.

English and Português do Brasil are first-class today and selectable per recording. Whisper itself understands 99 languages, so the rest are a settings toggle away. Summaries and chat answer in the language that was spoken, or in one you pin.

Through the system-audio tap built into macOS. No kernel extensions, no virtual audio drivers to install. You grant Microphone and System Audio Recording permission once, and both tracks are recorded separately so the transcript can tell you apart from everyone else.

Only if you ask it to. Meeting detection has three modes: stay out of the way, ask when a meeting app takes the microphone, or start automatically. The default is to ask. The menu-bar icon carries a red dot whenever a recording is running.

You decide. Settings has an allowlist of apps that trigger detection, so Zoom, Teams, Meet, Slack, WhatsApp or anything else can be included or ignored individually.

That depends on where you and the other participants are. Many places require everyone's consent. Tape does not ask for you: telling the room you are recording is your call to make.

Good enough that you read the transcript instead of relistening, and imperfect enough that you would not paste it into a contract. Accents, cross-talk and bad microphones cost accuracy. A larger Whisper size helps, at the cost of speed. Custom vocabulary in Settings fixes the spelling of names, acronyms and product terms.

It separates your microphone from everything the Mac played, so you get you against everyone else. It does not yet identify who is who inside the second track.

Answer questions about the meeting, draft the follow-up email, pull out what was assigned to you, and explain a specific moment if you right-click a transcript line and send it over. It reads the transcript, not the audio.

As much as fits. Tape keeps the opening and the ending and drops the middle when a transcript exceeds the context budget, which you control with a slider in Settings. It tells you how many minutes of speech the current setting holds and how much memory that costs.

In ~/Documents/Tape/ by default, one plain folder per meeting holding the audio, the transcript, the summary and your comments. You can move that anywhere in Settings. See your files for the full layout.

The app is about 15 MB; the models are hundreds of megabytes to a few gigabytes. Bundling them would make every update enormous, so Tape fetches the ones you actually use, once. After that everything runs offline. The models page covers what each one is for.

Common on corporate proxies and filters, which often block Hugging Face where the models are hosted. You can download them from another network, or install them by hand. Step by step instructions: installing models on a restricted network.

Once the models are installed, yes. Recording, transcription, summaries and chat need nothing but your Mac. The update check simply fails quietly and the app carries on.

Tape checks this site shortly after launch and tells you if a newer version exists. It never installs anything by itself, and you can turn the check off entirely in Settings.