Introduction
Most meeting transcribers ship your audio to a server. The vendor summarizes it, indexes it, and keeps a copy in a bucket you cannot audit. That is fine for a public webinar. It is a problem for a board call, a legal review, a customer conversation under NDA, or a hiring debrief.
Trace">Trace takes the opposite approach. It runs entirely on your Mac, uses a local speech model, and never sends the audio anywhere. This review is for builders on Apple silicon who want to know whether the on-device tradeoff is worth it in 2026.
Key Features
Local transcription, no server round-trip
Trace uses a local speech model that runs on the Apple Neural Engine. Nothing about the meeting leaves the machine. There is no vendor account tied to your transcripts, no cloud storage, no processor to sign a DPA with. For confidential calls this is the reason to install it.
System and microphone audio, captured together
Trace runs as a menu-bar overlay and captures your mic plus system audio at the same time. It works with Zoom, Google Meet, Microsoft Teams, Slack huddles, and anything else that plays through the Mac. No bot needs to be invited into the call, which sidesteps the awkwardness of a third-party attendee showing up in a client meeting.
In-call moment flagging
Press ⌘K during a call and Trace pins a timestamped marker to the transcript. This is the most useful feature. When someone drops a commitment or an objection worth revisiting, the flag lets you find it in the transcript in one scroll rather than rereading 40 minutes of dialog.
Live transcript peek
⌘/ opens a live transcript overlay so you can catch a sentence you missed without asking the speaker to repeat themselves. Useful for accented English, dense technical dialog, and calls where you joined two minutes late.
Markdown output that fits real workflows
Transcripts and summaries export as Markdown with direct hooks into Obsidian, Bear, Drafts, Things, and Notion. If you already run a PKM setup, meeting notes land where the rest of your notes live rather than in a proprietary vault you have to log into.
Pricing Breakdown
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Local transcription, Markdown export, basic meeting summaries |
| Pro | See App Store | Unlimited transcripts, integrations with Obsidian, Bear, Drafts, and Things, plus the tracecli command-line tool |
The pricing page does not surface a number. You have to open the App Store listing to see what Pro actually costs, which is a small annoyance when you are evaluating tools side by side.
How it compares on cost model
Cloud transcribers charge per minute or per seat per month, so the bill scales with how much you meet. Trace is a flat App Store purchase, so a heavy-meeting week costs the same as a light one. If you sit in more than a few hours of calls a week, the flat model wins on price alone, before you factor in the privacy story.
Pros and Cons
Pros
- Fully on-device. Transcripts and summaries never touch a server, and there is no vendor with a copy of the audio.
- Works as a menu-bar overlay with any conferencing app. No bot invites, no calendar integrations to configure, no admin permissions to negotiate with IT.
- Supports dozens of languages through the local speech model.
- Markdown output pastes cleanly into any PKM or note-taking workflow without cleanup.
- Keyboard-driven UX. ⌘K and ⌘/ mean you rarely leave the meeting window.
Cons
- Requires Apple silicon and macOS 14.4 or later. No Windows, no Intel Mac, no Linux. If your team is mixed, half of them cannot run it.
- Local processing is slower and less accurate than cloud models on long or noisy calls. A 90-minute all-hands with three people talking over each other will produce a transcript with more errors than a hosted service would.
- No mobile app and no cross-device sync. If you take a call on your iPhone, Trace is not in the loop.
- Pricing is behind an App Store click. You cannot compare it against alternatives without leaving the site.
- No shared workspace. Trace is single-user by design, which is coherent with the privacy story but means teams cannot pool transcripts through the app itself.
Who Is It For
Trace fits a specific reader. You are on an Apple silicon MacBook, you take calls that carry information you would rather not hand to a third-party processor, and you already write notes in Markdown. Solo consultants, lawyers, therapists, founders in due diligence, and engineers working under NDA are the clearest fits.
It is a bad fit if your team runs on Windows, if you need a shared transcript library across a company, or if you rely on the highest possible accuracy for long-form calls that get read closely later. Cloud tools still win on raw quality for those cases, and you should pay that cost knowingly.
Verdict
Trace earns a 6.6 out of 10. The on-device model is the point of the product, and it delivers on it in a way that a cloud tool with a privacy-flavored marketing page cannot. The keyboard UX, the Markdown output, and the menu-bar overlay are polished enough that it disappears into a working day rather than adding friction.
The ceiling is the platform. Apple silicon only, single-user, no mobile, and an accuracy tradeoff on long calls. Trace is honest about all of it, but those constraints are real and they narrow the audience.
If you fit the profile, Trace">install Trace and try it against your next confidential call. If you are on Windows, or you need team-shared transcripts, look elsewhere and revisit when your setup changes. That is the recommendation.
