Voice to Text for Medical Scribes:
Dictate EHR Notes in Real Time

The Scribe's Entire Job Is Throughput

A medical scribe's value to the physician is measured almost entirely in throughput. Can the chart be documented in real time, during the encounter, so the physician walks out of the room with the note essentially complete? Or does the physician end up doing two hours of after-hours charting because the scribe could not keep up?

Typing fast enough is the entire skill. A 90-words-per-minute typist on a good day produces a 150-word HPI in about 100 seconds, plus the cognitive lag of organizing what to type while the physician is talking. The reality is that complex encounters generate documentation faster than any human can type, which is why so much of the scribe role still ends up being post-encounter cleanup.

StarWhisper is a local Windows dictation app that closes that gap. The scribe speaks the HPI, ROS, exam, and assessment plan into the EHR field directly while the encounter is happening, at speaking speed. A 150-word HPI takes 45 seconds to dictate plus 10 seconds for the scribe to scan and approve. That is roughly 3x typing throughput, which is enough to keep up with most encounters in real time.

How Scribes Actually Use It in the Workflow

The flow most ED, urgent care, primary care, and specialty scribes settle into:

• Patient is roomed. Scribe opens the EHR chart, navigates to the HPI section.
• Physician begins the history. Scribe holds the hotkey and dictates the HPI in real time using the physician's spoken phrasing, lightly cleaned up for note format.
• For the exam, scribe uses the EHR's dot-phrase or template to insert the normal exam, then dictates the abnormal findings with the hotkey.
• For the assessment and plan, scribe dictates the physician's verbalized assessment directly, using template language for the plan steps where the EHR supports it.
• End of encounter: 80 to 95 percent of the note is already done. Scribe finishes the remaining sections (orders, billing codes, attestation pendings) in the few minutes after the physician walks out.

This is the same workflow that Dragon Medical scribes have used for years, but with a free-or-$10/month tool that does not require a per-seat enterprise license and that runs locally so PHI never leaves the workstation. The interaction model is the simplest possible: hold the key, talk, release.

For telehealth scribes, the flow is similar but the patient encounter happens in a video window (Zoom for Healthcare, Doxy, the EHR's own telehealth widget). The scribe still types into the same EHR text fields, just with a smaller view of the room. StarWhisper does not care; it writes to whatever has focus on Windows.

EHR Compatibility, in Detail

The question every scribe asks first is "does it work in Epic?" The answer is yes, because StarWhisper does not integrate with Epic or any other EHR specifically. It writes to the Windows text field that currently has focus, the same way a regular USB keyboard does. If the scribe can click into a note section and type, StarWhisper can dictate there.

If an EHR is not listed, the same rule applies: text field on Windows, StarWhisper dictates into it. Some hospital-managed thin-client setups (Citrix, VMware Horizon) require the dictation to happen in the local Windows session rather than the remote session, depending on how the EHR is accessed. Test once at the workstation type the scribe will be using.

Privacy and the PHI Question

The single biggest reason medical scribes pick StarWhisper over cloud-based alternatives is the PHI handling. Cloud dictation services upload audio (which contains PHI by definition during a patient encounter) to a third-party processor. That triggers a Business Associate Agreement, a security review by the hospital, and ongoing audit obligations. Many hospital IT departments refuse to allow it.

StarWhisper's Local Mode processes audio entirely on the workstation's CPU or NVIDIA GPU. Nothing is uploaded. The transcribed text is written into the EHR text field and only the EHR text field. There is no Business Associate relationship to establish because StarWhisper is not handling PHI; the audio is processed and discarded locally without ever touching a network. For most hospital IT reviews, this collapses the conversation: there is no cloud component to audit.

The HIPAA-compliant dictation software page covers the architecture details and the standard caveats. Short version: StarWhisper is designed to support HIPAA-sensitive workflows in Local Mode. Scribe agencies and hospital employers should review the architecture with their own compliance program, as is standard for any clinical software.

Cloud Mode is available but off by default. It should stay off for PHI work.

Why Dragon Medical Is Not the Default Anymore for Scribes

Dragon Medical One is the legacy industry standard, used by tens of thousands of physicians directly and by scribes who shadow them. It has deep EHR integrations, decades of medical-vocabulary tuning, and the enterprise sales relationships that put it on most hospital approved-software lists.

The problem for independent scribes and small scribe agencies is the pricing. Dragon Medical One runs at enterprise per-seat licensing that effectively requires a hospital contract. Individual scribes cannot buy a license. Scribe agencies pay per active user, and the cost is high enough that many smaller agencies don't equip their scribes with it at all, forcing the scribes to type.

StarWhisper at $10/month per seat (with a free tier that covers 500 words/day for testing) collapses that economics. A scribe can install it on their own workstation or their employer's workstation, use it for the entirety of a shift, and the cost is trivial. The accuracy is comparable for most clinical vocabulary; Whisper is not specifically tuned for medical content the way Dragon Medical was, but in practice the gap is smaller than the marketing claims and the local-processing advantage outweighs it for most scribe use cases. For physicians considering the same question, the voice to text for doctors page covers the comparison in more depth.

Realistic Time Savings on a Shift

The math, based on what scribes report after switching:

• Real-time HPI documentation: A typed 150-word HPI takes about 100 seconds, plus cognitive lag. Dictated, it takes 45 to 60 seconds total. Per patient, the scribe saves 40 to 55 seconds.
• Exam and A&P sections: Combined with the EHR's dot-phrase support for boilerplate, dictation typically saves another 60 to 90 seconds per encounter.
• End-of-shift cleanup: The big one. Real-time documentation means 80 to 95 percent of each note is done at the close of the encounter. Post-shift cleanup drops from 60 to 90 minutes per shift to 10 to 20 minutes.

For a busy ED scribe seeing 20 to 30 patients per shift, the cumulative effect is 45 to 75 minutes of recovered time per shift, plus the qualitative win of leaving work on time instead of staying late to finish charts. For primary care scribes seeing 15 to 25 patients per day, similar order of magnitude.

Plus the social-work-adjacent counterparts (case managers documenting in the same EHR, see the voice to text for social workers page) report the same general pattern: documentation time roughly thirds when the typing step becomes dictation.

Pricing in the Scribe Context

The free tier covers 500 words/day and 3,500/week. That covers a few patient encounters per day, useful for a part-time scribe or someone testing the tool. Full-time scribes documenting 15 to 30 patients per shift will exceed this within the first few hours.

StarWhisper Pro at $10 per month or $80 per year removes the word cap entirely, adds GPU acceleration (essential for real-time documentation on a busy shift), and unlocks priority cloud-fallback for the rare edge case. A scribe agency equipping 20 scribes pays $200/month, which is far less than a single Dragon Medical One seat would cost.

For comparison context, see the free Dragon NaturallySpeaking alternative page, which covers the broader Dragon-replacement landscape that StarWhisper fits into.