Zoom AI Companion Alternative: On-Device Notes, No Admin Toggle

You click the AI Companion button in a Zoom call and get the message every locked-down workspace knows: the feature is disabled for your account. Or you’re on Zoom’s free Basic plan, where the button never lights up in the first place. Or it works fine — and your client just asked whether their conversation is being processed in Zoom’s cloud, and you didn’t love saying yes. All three roads lead to the same search: a Zoom AI Companion alternative that produces the summary without the plan requirement, the admin toggle, or the cloud.

This post lays out what AI Companion actually requires, why the usual alternatives (bot notetakers) just relocate the problem, and how to get transcripts and structured summaries generated entirely on your own Mac instead.

What AI Companion actually requires

AI Companion’s pitch is “included free,” and that’s true as far as it goes. The full dependency list is longer:

• A paid Zoom plan. AI Companion comes with Pro and above at no extra cost. It is not available on the free Basic tier. (Zoom also sells it as a standalone add-on for people without a paid seat — the “free” part assumes you’re already paying for Zoom.)
• An admin who turned it on. AI Companion is enabled at the account level, feature by feature — meeting summaries, smart recording, chat compose are separate toggles. In a company workspace, that’s an IT decision, and plenty of IT departments have decided no, usually on data-governance grounds. There is no user-level override.
• The host’s account, not yours. Whether a summary exists for a given meeting is governed by the host’s side. Join a client’s Zoom as a guest and your own AI Companion entitlement is irrelevant — you get notes if their account produces them and shares them with you.
• Zoom’s cloud. Summaries are generated server-side. For internal standups, most people shrug. For client calls, interviews, and anything under NDA, “the meeting content transits the vendor’s cloud” is exactly the sentence that gets AI features disabled by admins in the first place.
• Zoom. The obvious one, but it bites daily: AI Companion doesn’t follow you into Teams when a customer sends a Teams link, or into Meet, Webex, or a Slack huddle. A notes tool that only covers one of the platforms you meet on isn’t a notes system; it’s a notes lottery.

None of this makes AI Companion bad software. It makes it conditional software — and every condition is decided by someone other than you.

Why bot notetakers don’t fix this

The first alternatives people try are the cloud notetakers: Otter, Fireflies, and their cousins. They remove the Zoom-plan dependency, but look at what they keep. A bot joins your meeting as a visible participant, the audio streams to the vendor’s servers, and processing happens in their cloud instead of Zoom’s. If the reason AI Companion was disabled in your org is data governance, a third-party bot is a step backwards — now the meeting content sits with a vendor your IT team never even evaluated, and everyone in the call watched “Notetaker Bot” arrive in the participant list.

There’s also a quota meter running: Otter’s free tier caps you at 300 transcription minutes a month and 30 per conversation, which one weekly client call exhausts. The full Otter comparison covers this in detail, but the short version is that bot notetakers answer “my plan doesn’t include AI Companion” and leave “I don’t want meeting audio in anyone’s cloud” exactly where it was.

What a Zoom AI Companion alternative needs to replace

Strip AI Companion down to what you actually use it for and there are three jobs:

1. A transcript of what was said, searchable later.
2. A structured summary — what was decided, who owes what.
3. Zero effort during the meeting — it runs while you participate.

All three jobs can run on the Mac that’s already playing the meeting. Two Apple frameworks do the heavy lifting:

• ScreenCaptureKit captures the system audio of a specific process — the Zoom app — scoped so that Music, notification pings, and the rest of your Mac stay out of the recording. macOS files this under the Screen & System Audio Recording permission, but it’s audio only: never video, never a pixel of your screen. Your microphone is mixed in, so the recording is the whole conversation — their side arrives through Zoom’s audio, your side through the mic.
• SpeechAnalyzer, Apple’s on-device speech framework, transcribes the stream live while the call runs. In MacStories’ independent testing (June 2025) it ran roughly 55% faster than Whisper Large V3 Turbo on-device on Apple silicon. No audio leaves the Mac.

Dictanta wires these together and adds the summary layer with Apple Foundation Models, also on-device: a TL;DR, action items with best-effort owner and due-date inference, decisions, and open questions. Every bullet is audio-anchored — click it and playback jumps to the moment it came from, so when the summary says “agreed to ship without the migration,” you can hear whether that’s what was actually agreed. That check is the part no cloud summary offers, and it’s what makes an on-device summary trustworthy enough to paste into a client email.

Nothing in this pipeline touches Zoom’s account settings. There’s no plan requirement, no admin toggle, no bot in the participant list, and no cloud on either Zoom’s side or anyone else’s.

The workflow, step by step

Assume a Mac running macOS 26.4 or later on Apple silicon (M1 or newer).

1. Join the Zoom call normally. Nothing changes on the Zoom side — no bot to invite, no setting to flip, no co-host permission to request. This works when you’re the host, a participant, or a guest on someone else’s locked-down enterprise account.

2. Start recording from the menu bar. Click Dictanta’s menu-bar icon and hit Record. Zoom is one of the apps Dictanta auto-detects — alongside Teams, Webex, Slack, and the major browsers — so the source picker’s This meeting option already points at Zoom’s audio with your mic mixed in. The first system-audio recording triggers a one-time macOS permission prompt; grant it once. If your calls start on short notice, enable the optional global shortcut in Settings and ⌃⌥⌘R starts a recording from anywhere.

3. Tell people you’re recording. AI Companion shows participants a disclosure when it’s active; a local recording shows nothing, so the disclosure is yours to make. “I’m recording this for notes — it stays on my Mac” covers it. Consent law for recorded calls varies by jurisdiction and applies no matter what tool you use.

4. Stop, and the notes are already there. When the call ends, stop the recording. The transcript is complete the moment you stop — it streamed live during the call — and the summary generates on-device right after: TL;DR, action items, decisions, open questions, each anchored to the audio. The recording titles itself with a timestamp, then auto-renames from the summary topic. The transcript autosaves every 30 seconds during the call, so a mid-meeting crash leaves a recovered recording, not nothing.

5. Share it your way. Export Markdown, PDF, Word (.docx), SRT, or JSON, or use Share Recap to produce a clean summary message for the meeting chat or a follow-up email. AI Companion posts its summary where Zoom decides; you paste yours where the team actually works.

How the options compare

	Zoom AI Companion	Bot notetakers (Otter, Fireflies)	Dictanta
Needs a paid Zoom plan	Yes (Pro+)	No	No
Needs admin enablement	Yes	No (but bots are often blocked)	No
Works as a guest on any account	No — host-side	Usually	Yes
Works outside Zoom	No	Zoom, Meet, Teams	Any app playing audio on the Mac
Where processing happens	Zoom’s cloud	Vendor’s cloud	On the Mac
Visible in the call	Disclosure notice	Bot in participant list	Nothing (disclosure is on you)
Speaker labels	Yes	Yes	No (planned)
Verifiable summary	No — text only	Partial	Every bullet audio-anchored

The speaker-labels row is worth dwelling on, because it’s the honest concession.

Where the local approach falls short

• No speaker attribution. Zoom’s cloud knows who’s on which audio line, so AI Companion’s transcripts name speakers. A local capture receives one mixed stream, so Dictanta’s transcript doesn’t tag who said what. Diarization is planned, but if per-speaker attribution is a hard requirement today, AI Companion genuinely does something the local path doesn’t.
• It’s Mac-only for meeting capture. System-audio capture requires the Mac frameworks. Take the call on your phone and there’s no Mac audio to capture — though the same recordings sync their transcripts and summaries across your Apple devices via your own iCloud. The audio itself never syncs anywhere.
• No automatic distribution. AI Companion can deliver its summary to participants on its own. Dictanta has no calendar integration, no webhook, no team workspace — sharing is a deliberate export or a pasted recap. For some teams the automation is the point; it’s not offered here.
• It records from when you press the button. No retroactive capture, and no meeting-ended auto-trigger. The global shortcut shrinks the gap to a keystroke.

When AI Companion is the right call

If your org pays for Zoom, your admin has AI Companion enabled, every meeting that matters happens inside Zoom, and cloud processing is acceptable to everyone on the call — use it. It’s included in the plan you’re already paying for, it attributes speakers by name, and it distributes notes without you lifting a finger.

The local alternative earns its keep when any of those conditions fail: you’re on Basic, the admin toggle is off, the meeting is on someone else’s account, half your calls happen on Teams and Meet anyway, or the meeting involves a client, candidate, or source whose audio shouldn’t transit any vendor’s cloud — the same reasoning that makes a private transcription setup worth having in general.

Trying it on your next Zoom call

1. Install Dictanta from the Mac App Store. macOS fetches the speech model once if it’s missing; transcription then runs fully offline.
2. Grant the microphone and system-audio permissions on first run.
3. Join your next Zoom call, hit Record, and say you’re recording.
4. When it ends, read the summary and click a bullet to hear the moment behind it.

The first 3 recordings are free with every feature unlocked and no length cap — enough to run it against AI Companion on real meetings and compare the notes side by side. After that it’s $9.99/month, $79.99/year, or $149.99 once for lifetime. Nothing is metered per minute, and there’s no account to create.

Bottom line

Zoom AI Companion is a good feature wrapped in conditions: a paid plan, an admin’s blessing, the host’s account, Zoom’s cloud, and Zoom itself. A Mac can deliver the same three jobs — transcript, structured summary, zero effort during the call — with no conditions attached, because everything runs on hardware you already own. The tradeoff is real but narrow: you give up speaker names and automatic distribution, and you gain independence from every toggle someone else controls, coverage of every meeting app on your Mac, and summaries you can audit against the audio they came from. If the AI Companion button has ever been grey when you needed it, that trade sells itself.

FAQ

Is Zoom AI Companion free?

It’s included at no extra cost with paid Zoom plans (Pro and above), but it isn’t available on the free Basic tier, and your account admin has to enable it before the button does anything. Zoom also sells AI Companion as a standalone add-on. If you’re on a free account or your admin has it switched off, the feature effectively doesn’t exist for you.

Can I use AI Companion if my admin disabled it?

No. AI Companion is controlled at the account level, feature by feature, so an individual user can’t override the admin setting. The practical alternative is to take notes on your side of the call instead: a Mac can record the meeting audio it’s already playing and transcribe it locally, which needs nothing from Zoom’s account settings at all.

What’s the best AI Companion alternative that doesn’t use the cloud?

An on-device recorder on your Mac. Dictanta captures Zoom’s audio plus your mic with no bot in the call, transcribes live with Apple’s on-device speech models, and summarizes with Apple Foundation Models — TL;DR, action items, decisions, and open questions, each linked to the audio. No audio leaves the Mac, and it works in any meeting, not just Zoom.