For teams & regulated practices

Dictation your compliance
review can actually approve.

Most dictation tools send your audio to a server. For a law firm, a clinic or a therapy practice, that single fact ends the evaluation — so a lot of professionals simply gave up on dictation and went back to typing. Whisper Master transcribes entirely on the Mac in front of you. There is no server to send audio to.

The reason this clears review is structural, not contractual

A cloud vendor asks you to trust a policy: they promise not to retain your recordings, and you take that on faith backed by a contract. The speech still leaves your building, so the risk is real and the paperwork is how you manage it.

Whisper Master runs the speech model on your own hardware. The audio is processed in memory on the Mac and never written to a network socket. There is no retention policy to audit because there is nothing to retain — a difference your reviewer can verify from the network layer rather than from our word.

What we will not claim: the app is not air-gapped. Signing in, downloading the model the first time, and checking for updates all touch the network. Product analytics and usage statistics are currently on by default. For a regulated deployment those are switched off and enforced — see the disclosure table below, which lists every byte that moves.

3.4%measured word-error rate on clean speech — and we publish the bad numbers too
0bytes of audio or transcript transmitted, ever
Offlinedictation keeps working with the network cable pulled out

Everything that touches the network

This is the whole list. Your reviewer will ask for it, so it is here rather than behind an NDA — and on a regulated plan the optional rows are disabled by policy, not by a checkbox a user can re-tick.

WhatWhere it goesOn a regulated plan

Your speech audioNowhere. Processed in memory on your Mac.N/A

The transcribed textNowhere. Pasted into your app, stored locally.N/A

Sign-in (identity only)Clerk — email address, no contentRequired for licensing

Speech model downloadCloudflare R2, once, ~1.5 GBOne-time; can be pre-staged by IT

Update checksSparkle appcastCan be pinned to an internal feed

Product analyticsPostHog and Google Analytics — account-linked, content-freeDisabled and enforced

Usage statisticsOur dashboard — counts only, no textDisabled and enforced

Notes & reminders syncYour account — this one carries dictated textDisabled and enforced

Nearby Macs (mesh)A paired Mac on your LAN — off by defaultDisabled and enforced

The same table, written for individuals, lives at /trust. We have never had two versions of it and do not intend to start.

Per-seat pricing

Published rather than “contact us”, because you should be able to size this before you talk to anyone. Annual billing; seats are reassignable when someone leaves.

Team

$19/seat/month

billed annually · minimum 3 seats

₹649 per seat/month, billed annually

Everything in Pro, for every seat
One invoice, central billing, reassignable seats
Shared custom vocabulary across the team
Analytics and usage sync switchable org-wide
Email support with a named human

For teams who want central billing, not a compliance posture

Practice

$39/seat/month

billed annually · minimum 3 seats

₹1,299 per seat/month, billed annually

Everything in Team
Regulated Mode — all telemetry off by enforced policy, not a setting
Signed data processing agreement (GDPR Art. 28 / DPDP)
Written architecture note for your security review
Managed deployment and pinned update channel
We answer your security questionnaire, in writing

For law, healthcare, therapy and finance · roughly 60% under Dragon

Enterprise

Let's talk

50+ seats, or procurement requirements

Let's talk 50+ seats, or procurement requirements

Everything in Practice
Volume pricing
MDM / configuration-profile deployment
Internal model and update mirror
Custom vocabulary built from your own terminology

No SOC 2 report today — ask us where that sits

For context: Dragon Medical One, the incumbent in clinical dictation, is licensed at roughly $99 per user per month and is cloud-backed. We are not pretending to match its clinical vocabulary yet — we are saying that if what you need is accurate, private dictation into the app you already use, you should not be paying that.

Where we are not a fit yet

You would find all of this in week two of an evaluation. Finding it in minute two is better for both of us.

macOS on Apple Silicon only. No Windows, no iPad, no Intel Macs. If your practice is on Windows we are not a candidate and no roadmap promise will change that this year.
No SOC 2 report. We are a very small company. If your procurement process requires SOC 2 Type II before signature, we will fail it today — tell us anyway, because knowing which deals it blocks is how it gets prioritised.
Accuracy degrades on Bluetooth headsets and in noisy rooms — 16.3% and 23.5% word-error rate respectively, versus 3.4% on clean speech. Those are our own published numbers. A wired or built-in mic is a real requirement, not a nicety.
Centralised admin controls are early. MDM deployment, managed preferences and an org-wide policy lock are built for regulated plans on request rather than available self-serve. Ask, and we will tell you honestly what exists.

Tell us what your review requires

There is no sales team — this reaches the person who wrote the app. If we do not clear your bar, we would rather say so in the first reply than in the fourth meeting.