A free companion for PowerScribe®
Dictum runs beside PowerScribe and upgrades it in three ways. It installs in five minutes with no IT ticket, and nothing you dictate ever leaves your workstation.
Dictation that never drifts. Press F9 (or the PowerMic button already in your hand) and a modern offline speech engine puts the words straight into the open report. No voice profile to train, corrupt, or retrain: the same sentence produces the same text, every time.
Impressions draft themselves. Press F10 and an on-device medical AI drafts the IMPRESSION from the findings you just dictated. Your own words, placed in the report as a draft for your review.
Autocomplete that knows your reports. Dictum learns your phrasing and finishes your sentences right in PowerScribe. It imports your own AutoText in one click.
Free during the pilot · built by Aaron Abajian, MD, a practicing radiologist · 64-bit Windows 10/11 · no admin rights needed
Simulated demonstration. You keep working in PowerScribe; Dictum is the small always-on-top panel. Text enters the report only when you press the button.
Still on PowerScribe 360? Microsoft is retiring it: maintenance ends August 2026, end-of-life August 2027. The upgrade means new subscriptions and retraining your voice profile from scratch.
Dictum works with your PowerScribe 360 exactly as it is. Nothing to migrate, nobody to ask: modern offline dictation, autocomplete, and AI impressions on the system you already have. Today, free. Get Dictum →
Sound familiar?
“I dictated the exact same sentence I say twenty times a day. Different text. Again.”
Press F9 (or the PowerMic button already in your hand) and the words land straight in the open report. The engine is a modern transformer (Parakeet-TDT) with no voice profile to train, drift, or corrupt: the same utterance produces the same text. It also decodes about 10× faster than you speak on an ordinary CPU, so you'll never watch the recognizer catch up.
“It still doesn't know the words I use every single day.”
Start a word in PowerScribe and a suggestion popup follows your caret: Tab accepts, Esc dismisses. Dictum quietly learns your vocabulary and phrasing from the reports you write, locally, on disk you control, so after a few days your own sentences finish themselves. Import your own PowerScribe AutoText in one click; a curated radiology corpus covers the rest.
“I just spent five minutes describing it. Now I get to summarize it all over again.”
Press F10 and an on-device medical language model drafts the IMPRESSION from the findings you just dictated: a diagnostic synthesis of your own words, never a finding you didn't make and never an unsolicited recommendation. It lands in the report as a draft for your review; you edit and sign exactly as you do today.
PowerScribe and Dragon stay exactly as they are. Every Dictum feature is a checkbox: untick them all and you're back to stock in one second.
The record button on your existing PowerMic drives Dictum, press-and-hold or toggle, your choice. No new hardware on the desk.
A compact, always-on-top panel in a warm, low-glare theme that sits politely next to your PACS monitors and stays out of the way.
The problem is documented, not anecdotal: in peer-reviewed studies, 23% of speech-recognition radiology reports contained at least one major error, versus 4% with transcriptionists (AJR, 2011). Dictum builds on its author's commentary in Radiology: Artificial Intelligence, “No Knew Typos: Modernizing Radiology Dictation” (RSNA, 2022), on using transformer language models to catch exactly these errors.
bytes of PHI leave your machine. No cloud, no login, no telemetry. The speech model, the impression model, and everything the app learns live on your disk.
file to install. Download, double-click, done. No installer wizard, no admin rights, no IT ticket.
changes to your sign-off. PowerScribe remains the system of record. Dictum only ever writes into the report on your explicit action: a button you press, a key you hit.
For you, and for your IT team
The installer and app are Authenticode-signed (Microsoft Azure Trusted Signing), and every release publishes its SHA-256 in latest.json. Updates verify both before installing.
Installs to %LOCALAPPDATA%\Dictum like Chrome or Zoom user
installs: no admin rights, no services, no drivers, no system changes.
Uninstalls cleanly from Windows Settings → Apps.
Two things: downloading the app and its AI models (from dictum.report, GitHub, Hugging Face, nuget.org), and a once-daily version check. Dictation audio and report text are processed on the workstation and are never transmitted. The app has no account, no telemetry, no analytics.
A business-associate agreement covers vendors that receive PHI. We receive none, by architecture: there is no server that could see your data. The privacy statement is the reference document for your compliance office.
Coming soon
The same engine, grown into a complete report editor: templates, tab-navigable fields, voice-driven picklists, and crash-proof drafts, all speaking PowerScribe's native report format. It stays in live two-way sync with the open report, and finished reports land back in PowerScribe with their fields still live, so you keep signing exactly where you sign today.
Simulated preview of the upcoming editor. Sync runs both ways: edits in either window mirror to the other, and writing into PowerScribe only happens when you say so.
Authenticode-signed · SHA-256 (Dictum.exe inside the zip):
published in downloads/latest.json
Dictum installs for your user account only, the same way Chrome or Zoom install: nothing system-wide, no IT ticket. On first launch it fetches the app and offers a one-time download of its on-device AI models (~3 GB). The speech and impression AI live entirely on your workstation, so nothing you dictate ever leaves it. No BAA, no legal review, no data-sharing agreement, because there's no data to share. And every feature is a checkbox: untick them and PowerScribe is exactly as you left it.
Advanced: standalone full exe · offline zip (~3 GB, models included)
Hospital network blocking this page or the download? Use the mirror: dictum-report.github.io
Dictum.zip and run Dictum.exe inside. It sets itself up in seconds, starts, and finds your open PowerScribe report by itself.