Medical dictation software that keeps PHI on your device
Whisper is HIPAA-minded medical transcription software that runs 100% on the clinician's machine. Dictate notes straight into your EHR — no cloud, no subprocessors, no patient audio ever leaving the exam room. For doctors and practices, that's not a feature. It's the whole point.
PHI stays on-device
The speech model runs on your own CPU/GPU. Patient audio and clinical notes stay on the machine that recorded them.
No subprocessors
No servers, no third-party AI scribe in the path — so there's no cloud vendor for your practice to vet or sign a BAA with.
Works offline
After a one-time setup, Whisper needs no internet. Suitable for locked-down clinic networks and shared workstations.
The problem with cloud medical dictation
Most cloud medical dictation software and AI scribes work the same way under the hood: they stream a patient's voice to a remote server, run it through third-party AI, and store the transcript somewhere you don't control. Every one of those hops is Protected Health Information leaving your practice — and a new party you have to trust, audit, and paper over with a Business Associate Agreement.
Whisper takes the cloud out of the equation entirely. The transcription happens on the same device where you dictate, so the audio and the note never travel anywhere. There is no upload to intercept, no server to breach, and no subprocessor list to review — because none of it exists.
| Data | Leaves your device? | Detail |
|---|---|---|
| Patient audio / dictation | Never | Transcribed on-device; never uploaded, never stored by us |
| Transcribed clinical notes | Never | Stay on the clinician's machine, typed straight into your EHR |
| License key | Key check (no content) | Key and device ID at activation, then about once a day when online. No content |
| Speech model files | One-time download | One-time download of the on-device model from our CDN |
| Update checks | Optional | A version check you can ignore or block on clinic machines |
Because PHI never leaves the device, Whisper removes the biggest exposure in most dictation workflows — cloud transmission and storage — and supports HIPAA-minded practices without a cloud BAA to negotiate.
On-device dictation software for doctors
Whisper installs like any desktop app and runs the AI model directly on your hardware. That single design choice is what makes it a genuinely private medical dictation tool — and it comes with practical benefits for a busy clinic.
- PHI stays on the machine that recorded it — nothing to intercept, nothing to breach in transit.
- No cloud account, no admin console, and no subprocessor to add to your risk register.
- Runs on the computers you already have — no dedicated dictation hardware or on-prem server.
- Works on shared and locked-down clinic workstations, fully offline after setup.
We're transparent: Whisper doesn't carry a HIPAA or SOC 2 certification badge. Instead it gives you an architecture where the sensitive data never leaves your control — which is what most security reviews are really trying to guarantee.
Dictate into any EHR — and every other app
Whisper types wherever your cursor is, so it drops text straight into your existing charting workflow. There's no integration to build and no plugin to keep updated — if a field accepts typing, it accepts your voice.
Any EHR / EMR
Epic, Cerner, athenahealth, eClinicalWorks, or any browser-based system — dictate into the note field directly.
Every other app
Referral letters in Word, messages in secure email, entries in your practice-management tool. One dictation tool for all of it.
Automatic formatting
Punctuation and capitalization are added for you, so notes read cleanly without manual clean-up.
Fast enough to keep up with a full clinic day
On-device doesn't mean slow. Whisper is GPU-accelerated — Metal on Mac, Vulkan on Windows and Linux — and transcribes roughly an hour of audio in about a minute on Apple Silicon. Notes are ready almost as fast as you can speak them, so dictation keeps pace with back-to-back patients instead of piling up for later.
~1 min per hour of audio
Roughly an hour of dictation transcribed in about a minute on Apple Silicon.
GPU-accelerated
Uses Metal on Mac and Vulkan on Windows and Linux to stay fast on existing clinic hardware.
No queue, no upload wait
Processing is local, so there's no round-trip to a server between speaking and seeing text.
Yearly pricing for practices — no subscription
Dragon Medical One and cloud AI scribes bill every clinician every month, indefinitely. Whisper is a yearly license with no auto-renewal: $39 for the first seat and $29 for each additional seat per year, for teams of up to 100. License the whole practice with one license key, activated on each provider's device up to your seat count. No card is charged again: you decide each year whether to renew.
$39 first seat
Per year. No subscription, no per-note fee, no auto-renewal. Renew only if you want to.
+$29 per additional seat
Per year. Add providers and support staff at a flat per-seat price, for teams of up to 100 clinicians.
Want to test it first? Request a free 14-day pilot for up to 5 devices.
Rolling out across a larger group? See Whisper for teams and business for volume licensing details.
Medical dictation software FAQ
Is Whisper HIPAA compliant medical dictation software?
Compliance is a property of your whole workflow, not a single tool — but Whisper is built to support HIPAA-minded practices. Because Protected Health Information (the audio and its transcript) never leaves your device, Whisper removes the biggest exposure in most dictation workflows: transmitting and storing PHI in the cloud. There is no cloud vendor to sign a BAA with, because there is no cloud. We are transparent that Whisper does not carry a HIPAA or SOC 2 certification badge; instead it gives you an architecture where the sensitive data simply stays with you.
Does any patient data or PHI leave my device?
No. All medical transcription runs locally on your own machine using an on-device AI model. Your dictated audio and the resulting transcript are never uploaded to our servers or to any third party. There are no subprocessors in the transcription path. This is the core difference from cloud dictation and AI-scribe tools, which route a patient's voice and notes through their servers.
Does it work with my EHR or EMR system?
Yes. Whisper types text wherever your cursor is, so it works inside any EHR, EMR, or web-based charting system — Epic, Cerner, athenahealth, eClinicalWorks, or any other — as well as email, Word, and every other application. There is no integration to install and no plugin to maintain; if you can type into a field, you can dictate into it.
How much does it cost compared to Dragon Medical?
Dragon Medical One is a subscription that typically runs several hundred dollars per clinician per year, and cloud AI scribes bill monthly on top of that. Whisper is a yearly license: $39 for the first seat and $29 for each additional seat per year, for teams of up to 100. There is no subscription, no auto-renewal, and no per-note fee. You renew each year only if you want to.
How accurate is the medical transcription?
Whisper uses a modern speech-to-text model with automatic punctuation and capitalization, and it is fast — it transcribes roughly an hour of audio in about a minute on Apple Silicon. Accuracy is strong on natural clinical dictation. As with any dictation software, we recommend a quick read-through before signing a note, especially for drug names and dosages.
Which platforms and devices does it run on?
Whisper runs natively on macOS (Apple Silicon), Windows 10 and 11, and Linux. It is GPU-accelerated on each — Metal on Mac, Vulkan on Windows and Linux — so transcription stays fast on the hardware you already have in the clinic. After a one-time setup it works fully offline.
Talk to us about your team
Pilot, invoice, purchase order or a security questionnaire. We reply by email, usually within two business days.
Private medical dictation your compliance team will approve
On-device, offline, no subprocessors — PHI never leaves the exam room. A yearly license from $39, with per-seat pricing for the whole practice.
Get Whisper — from $39/year