EHR Dictation Without the Citrix Headache (Windows & Mac)

Ask a clinician what stands between them and dictated notes, and the answer usually isn't accuracy — it's plumbing. The EHR lives behind Citrix or a browser tab, IT owns what gets installed where, and the classic dictation stack wants audio piped across the network into a virtual desktop before a word gets transcribed. So the notes get typed. At 10pm.

TL;DR: There are two architectures for EHR dictation, and picking the right one dissolves most of the pain. In-session (the Dragon Medical One path) transcribes inside the EHR environment — powerful, but through Citrix it requires a special client audio extension, per-machine installs, and IT-managed audio routing. Client-side (the approach this guide sets up) transcribes on the computer physically in front of you and types the finished text into whatever window has focus — a web EHR in Chrome or Edge, a Citrix session, a native app. Audio never crosses the network, nothing installs server-side, and it works on Windows and Mac. With Voibe, transcription is fully on-device on Apple Silicon Macs and runs through a private zero-retention cloud on Windows — and a custom Dictionary keeps drug names and clinical terms spelled right.

How you access the chart determines which dictation problems you inherit:

Two of the three are trivial. The middle one — the hospital standard — is where dictation projects go to die, because the traditional architecture insists transcription happen inside the session, which drags your microphone across the network with it. That's a solvable problem (Nuance solves it, with effort), but it's also an optional problem: move the transcription to the client side and it disappears.

Dragon Medical One (DMO) is the incumbent for a reason: medical vocabularies out of the box, voice commands that navigate EHR fields, and enterprise administration. It runs inside the EHR environment — and in a virtual desktop, that's precisely what creates the plumbing. Per Microsoft's DMO Citrix deployment documentation:

• Your microphone audio must reach the Citrix server, which means either USB redirection or Nuance's custom audio channels via the Dragon Citrix Client Audio Extension, installed on each client PC.
• The custom channels exist because native audio is heavy: they reduce dictation bandwidth to roughly 28 kbit/s per user, versus up to 1.4 Mbit/s for native audio channels.
• USB redirection can't be combined with the extension, and Microsoft's documentation recommends against relying on it — it "depends on the state of the network and is therefore unreliable." Troubleshooting audio in these setups is its own documentation section, and multi-hop environments get their own configuration guide.

None of this is a knock on DMO's engineering — it's what solving audio-into-VDI properly looks like. But it explains two realities: dictation this way is an IT project, not an app install, and it's priced like one. Nuance doesn't publish DMO pricing; reseller-quoted rates (verified August 2026) run $79-99 per user per month by contract term plus a $525 per-user setup fee — roughly $948-1,188 per year, before the health-system discounts large buyers negotiate. Our DMO cost breakdown has the full arithmetic, and DMO on Mac covers the browser-based route for Mac clinicians.

The client-side approach inverts the problem. Instead of shipping audio to where the EHR runs, transcription happens on the machine physically in front of you, and the finished text is inserted at your cursor — which can be a note field in a web EHR, a Citrix session window, or a native app. The session doesn't know dictation exists; it just receives text, like typing. Setup, using Voibe — ours — as the example:

1. Install on your local machine — the Windows PC or Mac you physically use, not the virtual desktop. Windows: native installer from getvoibe.com. Mac: the .dmg; on Apple Silicon (macOS 13+) a local Whisper model downloads once and transcription runs fully on-device. 7-day free trial, no account.
2. Grant permissions. Microphone (both platforms); on macOS also Accessibility, which is what lets it type into other windows.
3. Load the Dictionary with your clinical vocabulary. Drug names, procedures, referring physicians, local facilities. Entries bias transcription itself, so hydrochlorothiazide and Dr. Nguyen come out right from then on.
4. Click into the note field and dictate. Hold the hotkey, speak the narrative, release — the text appears at the cursor. In a web EHR that's the browser; through Citrix it's the session window; in a native app it's the field itself.
5. Verify insertion in your exact environment during the trial. Remote-display clients vary in how they accept synthetic input; most accept system-level keystrokes exactly like typing, but the honest advice is to test your Citrix/Horizon/RDP client for two minutes before you rely on it. (Web and native EHRs have no such variable.)

What you've avoided: the audio extension, the per-client IT deployment, the bandwidth question, and the server-side install. What you've kept: your voice, your vocabulary, and a tool that also works in your email, referral letters, and everything else you write — see our doctors' dictation roundup for how clinicians use that beyond the chart.

If your EHR runs in a local browser tab — ChartPath, athenahealth, and the growing class of cloud EHRs — there is no remote-desktop variable at all. The note field is an ordinary web text area; a system-wide dictation tool types into it exactly as it types into Gmail. Click into the HPI or narrative field, hold, speak the paragraph, release, and move to the next field.

Clinical narrative is also where dictation's speed advantage is largest: notes are long-form prose, often over a hundred words per entry, and speech at 150+ words a minute against roughly 40 typed turns a charting backlog into a talking task. Therapists and psychiatrists — whose notes are almost entirely narrative — get the same win, covered in our therapist dictation guide.

One category distinction worth keeping sharp: dictation types what you say into the chart. Ambient AI scribes (Sunoh and its peers) listen to the patient encounter and draft the note for you — a different tool with different consent, accuracy, and data-flow questions, compared in our AI medical scribe roundup and the broader dictation vs. notetaker taxonomy.

Where the audio goes is an architecture fact; whether a tool is permitted for your charts is a compliance decision. Both halves, plainly:

• Architecture. With client-side on-device transcription (Voibe on Apple Silicon Macs), audio containing PHI is processed on your machine and never leaves it — there is no vendor server in the path at all. On Windows, Voibe transcribes through its private zero-retention cloud running open-source models: audio is never stored, sold, or used to train AI. In-session tools process audio within your organization's managed environment under its enterprise agreements.
• Compliance. HIPAA compliance attaches to organizations, agreements, and practices — not to an app. We make no compliance claims for Voibe; your compliance officer weighs the architecture, your workflows, and your obligations. What the architecture buys you is a shorter conversation: "audio never leaves the laptop" is a much easier sentence to evaluate than a cloud vendor's data-flow diagram.

Our HIPAA dictation guide walks through the questions to bring to that conversation, and why offline dictation matters covers the architectural argument in depth.

Client-side dictation replaces the typing, not the whole DMO feature set. Choose in-session DMO when:

• Your organization already deploys and pays for it. The plumbing is IT's problem and solved; use it.
• You navigate the EHR by voice. DMO's commands (jump fields, invoke templates, sign) are real workflow features a text-typing tool doesn't replicate.
• You want medical vocabulary with zero setup. DMO ships specialty vocabularies; client-side tools get there via your Dictionary, which takes a week of adding terms as you go.

Choose client-side when you chart through a web EHR or remote session, IT isn't deploying anything for you, or you're an individual clinician or small practice paying your own way. The cost gap is not subtle: DMO's reseller-quoted $948-1,188/year plus $525 setup totals roughly $3,369-4,089 over three years; Voibe is $149 once (or $7.50/month) — about 4% of that. For clinicians displaced by the Dragon Medical Practice Edition sunset, that math is the whole story.

Text isn't appearing in the Citrix or remote window

First prove the tool works locally: dictate into Notepad or TextEdit. If local works but the remote window doesn't, click directly into the remote text field (focus must be inside the session), and check your remote client's keyboard/input settings — most accept system-level keystrokes by default. The insertion method matters here: tools that type the transcript as simulated keystrokes work even in locked-down sessions, while tools that insert via clipboard paste fail wherever clipboard redirection is disabled — which enterprise Windows desktops routinely do. If your setup allows clipboard redirection, dictating into a local scratchpad and pasting remains a serviceable last resort.

Drug names and clinical terms come out wrong

Add them to the Dictionary as they occur; each entry is permanent. Week one involves a dozen additions; week three, almost none. Generic speech models simply don't know apixaban until told.

The web EHR field behaves oddly with inserted text

Some structured fields (dropdowns, coded entries) only accept picks, not free text — dictation is for the narrative fields. If a narrative field validates on every keystroke and hiccups on a full inserted sentence, dictate in shorter phrases.

Wrong microphone in a shared workspace

Check the input device after docking or moving rooms — the dictation app captures whatever mic the OS hands it. A wired headset beats the webcam mic in a noisy clinic, and our microphone guide has specific picks.

Basics

Can I dictate into any EHR?
Any EHR with a text field you can click into — web, Citrix/remote, or native — accepts client-side dictation, because the text arrives as keystrokes. In-session tools support specific EHR environments by deployment.

Do I need my IT department's help?
For client-side dictation on your own machine, no server-side installation is involved — though your organization's software policies still apply to your workstation. For in-session DMO through Citrix, yes: extensions and audio routing are admin-deployed by design.

Citrix & remote

Why does Dragon need a special extension for Citrix?
Because it transcribes inside the session, so your audio must cross the network — the extension compresses dictation audio to ~28 kbit/s versus up to 1.4 Mbit/s native. Client-side tools transcribe before the network and need none of it.

Does client-side dictation add lag through Citrix?
Transcription happens locally, so the remote session receives plain text — the same path your typing takes, subject to the same session latency and nothing more.

Privacy

Where does patient-identifying audio go?
On-device mode: nowhere — it's processed on your machine. Voibe's Windows mode: a private zero-retention cloud; audio is never stored, sold, or used to train AI. Whether either fits your obligations is your compliance team's call — see our HIPAA dictation guide.

Cost

What does EHR dictation cost in 2026?
Dragon Medical One: reseller-quoted $79-99/user/month plus $525 setup (~$948-1,188/year; Nuance publishes no list price). Client-side: Voibe at $149 lifetime or $7.50/month. Three-year totals: roughly $3,369-4,089 versus $149.

The dictation problem in medicine was never the speech recognition — it's been the architecture. If your organization deploys Dragon Medical One, you're covered. If you're the clinician charting through a browser tab or a Citrix window with no IT project in sight, put the transcription on your side of the glass: install Voibe on the machine in front of you, load your clinical vocabulary, run the two-minute test in your EHR, and dictate tonight's notes instead of typing them. Download the free 7-day trial — on-device on Apple Silicon Macs, native on Windows, no account required.

Related reading:

• Best dictation software for doctors — the full clinical roundup
• HIPAA and dictation — the questions for your compliance team
• Dragon Medical One cost — the reseller-quoted math in detail
• AI medical scribes — the ambient-documentation category, compared
• Dictation for radiologists — the PowerScribe transition
• Voibe for Windows — the native Windows app, explained
• Best Windows dictation apps — the wider Windows field, ranked