Voice should be as ordinary as a keyboard
Nobody pays a monthly fee to keep their keyboard working. Dictation earned its place as a basic input method, so we priced it like one.
By The RadMyk team
There is a quiet absurdity in modern radiology software. The keyboard on your desk cost you nothing after the day you bought the computer. It does not phone home. It does not expire. It does not ask you to renew a license every January. You press a key, a letter appears, and that contract between you and the machine never changes.
Voice never got the same deal.
Somewhere along the way, dictation became a service: metered, subscribed, and rationed. A radiologist who speaks ten thousand words into a report has to keep a credit card on file to keep doing it. That is strange. Speaking is not a premium feature. It is the most natural input a human has.
The keyboard test
We built RadMyk around a single test: would this be acceptable for a keyboard?
- A keyboard you buy once. RadMyk is one-time pay.
- A keyboard keeps working if your internet drops. RadMyk keeps working.
- A keyboard does not read your documents back to a server. Neither does RadMyk.
- A keyboard does not care how much you type. Speak as much as you want.
If the answer for a keyboard is “of course,” then the answer for voice should be “of course” too.
The test is useful because it strips away software habit. We have become used to renting tools that sit between a clinician and the record. We have become used to cloud processing as the default answer. We have become used to every workstation needing one more account, policy, license, and renewal.
But radiology dictation is not magic. It is input. The value is not that a vendor hears you speak. The value is that the words land correctly in the report.
Abundance changes behavior
When something is scarce, you ration it. Radiologists who pay per-seat, per-month, per-word learn to dictate defensively. When something is abundant, like the keys under your fingers, you stop thinking about the meter and work.
That is the whole goal. We want voice input to disappear into the background of the reading room, the same way the keyboard already has. Not a tool you manage. A way you talk to the computer.
That has practical consequences. If the network fails, dictation should keep going. If you move from an EHR field to a PACS/RIS report box, speech should follow the cursor. If you work on a Mac at home and Windows in the reading room, the input method should not force a workaround. If you dictate more reports this month than last month, the tool should not meter your words.
This is why RadMyk runs on-device. Voice does not need to leave the room for text to appear. A local model can listen, transcribe, and paste into the active field. Privacy becomes part of the machine’s behavior, not a contract clause you hope is enforced somewhere else.
Ordinary is the point
Most software wants to feel important. Input tools should want the opposite. A good keyboard is not memorable. A good mouse does not announce itself. A good microphone does not ask to be admired. It works, and the work moves through it.
That is the standard voice should meet in radiology. Not another subscription to manage. Not another vendor portal. Not another cloud dependency between the radiologist and the report.
The mantra fits on a keycap: voice-based input should be as abundant and available as a keyboard.