An Otter.ai alternative with no bot and no minute limits

The bot is the thing people actually hate

Otter works by joining your meeting. It appears in Zoom, Google Meet and Microsoft Teams as a participant, records from inside the call, and produces a transcript from there.

That design has consequences you feel every week. Someone has to let it in from the waiting room. External guests see an account they do not recognise and ask about it. A client on a sensitive call notices a recorder in the attendee list and the temperature of the conversation changes. If the bot fails to join, you have no recording at all and you find out afterwards.

TNN never joins anything. It captures the audio your Mac is already playing, so there is nothing in the participant list, no meeting link to hand over, no waiting room, and nothing for the other side to notice. It works identically on Zoom, Meet, Teams, a Slack huddle, a phone call on speaker, or two people talking in a room. If your Mac can hear it, TNN can transcribe it.

The meter is the other thing

On Otter's Pro plan you get 1,200 transcription minutes a month, a 90-minute ceiling on any single conversation, and 10 file imports a month. Twenty hours sounds generous until you have a week of workshops, or one three-hour planning session that gets cut off at ninety minutes.

Business lifts the meeting transcription limit and raises the per-conversation cap to four hours, at $30 per user per month billed monthly or $19.99 billed annually.

TNN has no minute limit, and the reason is structural rather than generous: transcription runs on your Neural Engine, so a minute of audio costs us nothing. There is no meter to build.

Otter.ai pricing and limits checked July 2026 against public pricing pages. Vendors change plans often - verify before you decide.

What happens to the audio

Otter uploads and processes your meetings in its cloud. That is a normal SaaS arrangement and for a lot of meetings it is genuinely fine.

It is not fine when the meeting is a candidate interview, a compensation conversation, an unannounced incident, a client under an NDA that enumerates subprocessors, or anything a security questionnaire will ask about later. In those cases the honest answer to "where does the audio go" needs to be "nowhere".

With TNN that is the actual answer. Recording and transcription happen entirely on your Mac. There is no server, so there is nothing to breach, subpoena or add to a subprocessor list. The one online step is optional: AI-written notes run through your own Claude account, and you can see exactly which files Claude reads.

What you get that Otter does not have at all

Every TNN meeting is one plain Markdown file: your live notes, the transcript and the AI summary together, in a folder you choose. Delete the app and the files remain, readable in any editor.

Because they are plain files, TNN can expose them to Claude over MCP. Claude lists your meetings, reads notes and transcripts, runs skills you have saved, and writes results back into the meeting file. Ask what you committed to across the last month of client calls and the answer comes from transcripts rather than memory. That runs on your own Claude Pro or Max subscription, so there is no API key and no AI surcharge from us.

Otter gives you a searchable cloud archive. TNN gives you a folder your AI can reason over.

Where Otter is still the better choice

• You are not on Apple Silicon. TNN needs macOS 13 or later on an Apple Silicon Mac. Otter runs on web, Windows, macOS, iOS and Android.
• You need live captions. Otter's real-time captioning is a genuine accessibility feature and TNN does not match it.
• You need a team workspace. Shared folders, teammates searching each other's meetings, permissions. TNN is single-user today.
• You need to record meetings you are not in. A bot can sit in a call you skipped. Nothing local can.
• You want a mature product. TNN is in beta. You will find rough edges.

Trying it

Download, grant the screen and microphone permissions macOS asks for, connect your calendar so TNN knows which meeting is starting, and press record. The transcription model downloads once, about 600 MB, and runs on the Neural Engine after that. Language detection is automatic, including meetings that switch language mid-sentence.