“I dictated the exact same sentence I say twenty times a day. Different text. Again.”
Say it the same, get the same
PowerScribe's recognizer leans on a per-user voice profile that drifts, degrades,
and eventually needs retraining. Dictum uses a modern transformer
speech model (Parakeet-TDT) with deterministic decoding: no profile to
train, no profile to corrupt — the same utterance produces the same
text, with punctuation and measurements formatted the way radiologists
write them. It decodes a five-second utterance in about half a second on an
ordinary CPU.
“AutoText was clearly designed by someone who never dictated a case.”
Templates that fill themselves out
Import your own PowerScribe AutoText in one click — or start from
the built-in template library. Every field is tab-navigable, picklists pop up with
the default preselected, and voice commands drive it all: say
“next field”, speak an option, keep dictating. Finished reports
paste back into PowerScribe with the fields still live.
“It still doesn't know the words I use every single day.”
Autocomplete that learns you
The app quietly learns your vocabulary and phrasing from the reports you sign —
locally, on disk you control. Type three letters and your own sentence finishes itself.
A curated radiology corpus covers the rest, ranked by the section you're writing in.
Live mirror
The open report mirrors into the plugin in real time — fields, lists, formatting —
and stays read-only unless you say otherwise. Dictation in PowerScribe always wins.
Crash-proof drafts
Work is autosaved every few seconds with full field fidelity. Kill the app, lose power,
reboot — your draft comes back with tab navigation intact.
Made for reading rooms
A warm, low-glare theme that sits politely next to your PACS monitors, resizable,
with your font and size remembered.