Medical Charting Dictation on Windows:
Local, Free, Whisper-Accurate

Charting time is the real cost

Documentation time is a major source of burden in clinical work. Voice dictation can reduce typing load when accuracy and reliability are good enough for the clinician. When accuracy is poor or the tool is unreliable, providers fall back to typing and the time penalty stays.

The clinical dictation market has been dominated for years by Dragon (Nuance, now Microsoft), which serves the enterprise institutional buyer with Dragon Medical One: per-seat subscription pricing, specialty vocabulary packs, an integration framework, and enterprise admin tooling. For health systems with thousands of providers, that institutional package makes sense. For solo physicians, small groups, APPs working independently, locum tenens providers, telemedicine practices, and any clinician who would rather not navigate enterprise procurement to add a dictation tool to their workstation, the economics and the workflow do not always fit.

That is the gap StarWhisper is positioned for. It is a free local-first Windows dictation app powered by OpenAI Whisper. It is not Dragon Medical, and is not trying to be, but for everyday charting it covers the core need: turn what the provider says into accurate text in the EHR field, keep the audio on the workstation, and stay out of the way.

What Whisper accuracy looks like on medical content

OpenAI's Whisper is a broad multilingual speech recognition model. On medical content, the useful question is practical accuracy in your own environment: microphone, accent, model size, background noise, and specialty vocabulary all matter.

Where Whisper struggles is the long tail: unusual drug formulations, very obscure compounds, recent FDA approvals it has not seen in training, multi-syllable specialty jargon, abbreviations spoken as letters, and ICD-10 alphanumeric codes when dictated as a mixed letter-number sequence. Most of these are fixable in seconds. For everyday HPI, ROS, physical exam findings, A and P, telephone notes, and patient instructions, test a sample set before deciding whether the workflow saves time.

If you have used a free consumer dictation tool like the Win+H voice typing in Windows 11 and found it almost-but-not-quite adequate for charting, the gap between that and Whisper is meaningful. See the professional accuracy page for the broader accuracy discussion across use cases.

EHR compatibility, in concrete terms

Because StarWhisper places text through standard input or paste behavior at the OS layer, it can work in many EHR fields that accept typed or pasted text. There is no EHR-specific API integration to install. Examples to test include:

Always validate in your specific EHR build before using in production. For broader compatibility context across roles, see voice to text for doctors and voice to text for medical scribes.

SOAP and DAP-style structured notes

StarWhisper does not produce structured notes on its own; it produces accurate text. Structure comes from how you compose the dictation. A common workflow looks like this:

• Subjective: trigger your SOAP template SmartText for headers, then dictate the chief complaint, HPI, ROS, and pertinent history with StarWhisper.
• Objective: trigger normal exam SmartText, then dictate any pertinent abnormal findings, vitals interpretation, and lab observations.
• Assessment: dictate the differential, working diagnosis, and rationale.
• Plan: dictate the medication changes, follow-up, referrals, and patient education.

For behavioral health, the DAP structure (Data, Assessment, Plan) works similarly: dictate the variable narrative, lean on EHR templates for the structured fields. SimplePractice and TherapyNotes both have customizable progress note templates that pair well with this approach.

If you prefer a more linear narrative style, you can also just dictate the whole note from top to bottom in StarWhisper and clean up afterward. Both approaches work. The accuracy is high enough that cleanup is usually quick.

The PHI question, addressed directly

The most important architectural property of StarWhisper for clinical use is that, in Local Mode, transcription can run on the workstation without uploading audio for transcription. The Whisper model lives in the installation directory, runs on the local CPU or GPU, processes the audio in memory, and discards it after producing text.

For HIPAA-regulated practices, this can reduce the cloud-transmission risk category that drives many concerns about dictation software. It does not replace a practice-level HIPAA review. Clinical teams should verify workstation security, access controls, retention, update behavior, and Cloud Mode settings with their own compliance program. The detailed setup notes are on the HIPAA dictation page, including how IT can lock Cloud Mode off at install time for fleet deployments.

Cloud Mode sends audio to the selected cloud transcription provider and is not appropriate for PHI workflows unless the practice has separately approved that configuration. Practices deploying for clinical use should keep Cloud Mode disabled. For a HIPAA-sensitive deployment, follow the setup steps on the HIPAA page, including the administrator configuration option that hides the Cloud Mode toggle from end users.

Where Dragon Medical still wins, honestly

Several scenarios where Dragon Medical or another enterprise clinical dictation product remains the better choice, and we say so directly because matching tool to workflow matters more than winning a marketing argument.

• Specialty vocabulary packs. Radiology, pathology, cardiology, and a few others have heavy specialty vocabularies where Dragon's pre-built dictionaries genuinely save time.
• Voice command frameworks for structured documentation. Dragon's macro and "say X to insert Y" framework is mature, well integrated with Epic and Cerner, and harder to replicate.
• Ambient scribe. Nuance DAX and similar tools listen to the full encounter and draft a note. StarWhisper does not do that and is not pitched as a substitute.
• Enterprise admin tooling. Centralized user profile management, organization-wide analytics, SSO, integration with provider directories, all standard at the enterprise level.
• BAA and enterprise compliance documentation. Dragon Medical One comes with extensive HIPAA and SOC documentation that simplifies institutional procurement.

StarWhisper is positioned for the providers who do not have or want those wraparound services: solo practitioners, small groups, locum and telemedicine providers, APPs, and physicians frustrated with Dragon Medical's licensing complexity. The free tier exists so you can validate the workflow on your machine before any procurement conversation.

Setup for a practice or solo provider

For individual providers:

• Install from the download page or the Microsoft Store.
• Confirm Local Mode is enabled (default) in Settings, Transcription mode.
• Choose a push-to-talk hotkey that does not conflict with EHR shortcuts.
• Test in a non-PHI field first (a scratchpad in Notepad), then in your EHR with a test patient or training record.
• Confirm with your compliance program before charting on live PHI.

For a small practice rolling out across multiple providers, the same flow scales. There is no per-clinician voice training, no profile migration, and no per-user licensing dance. Pro licenses can be purchased individually or as a small batch. For deployments where IT manages the workstations, the administrator configuration option can lock Cloud Mode off at install time so providers cannot accidentally enable it.