What You'll Learn
- Why the after-hours note-writing grind is a voice-input problem, not a time-management problem
- The compliance difference between cloud dictation services and local voice dictation
- A template-driven SOAP workflow that turns a session into a note in 3–5 minutes
- Why the distinction between progress notes and psychotherapy notes matters for where you dictate
- The specific Mac setup therapists use with SimplePractice, TherapyNotes, TheraNest, and Valant
The Math of the Note-Writing Problem
A few years ago, one of our beta testers was a licensed clinical social worker in private practice. She told me the most honest thing anyone has said about therapist documentation: “I love the work. I resent the notes.”
She was seeing around thirty clients a week. Each session generated a progress note of 150–400 words depending on treatment modality, complexity, and whether insurance was involved. Typing those notes took her an average of ten minutes per client. That's roughly five hours a week — almost a full clinical day — spent on documentation. She was doing most of it at night, at home, after dinner, after her kids were in bed.
This is not unusual. Research published by the American Psychological Association has consistently found that clinical documentation is one of the top contributors to burnout among mental health clinicians. The problem isn't complexity; it's volume. The notes are short. There are just a lot of them.
Voice dictation changes the ratio. Speaking a 250-word note takes about 90 seconds at normal conversational speed. Add review, editing, and a moment to structure your thoughts, and you're at 3–5 minutes per note instead of 10–12. Across a week, that's roughly three clinical hours reclaimed.
Why Cloud Dictation Creates a Problem You Don't Want
The obvious solution to the note-writing problem is voice-to-text. The non-obvious part is that most voice-to-text tools solve the time problem by creating a privacy problem.
Consumer cloud dictation tools — browser extensions, general-purpose apps, mainstream speech-to-text services — stream your audio to a third-party server for processing. When that audio contains a client's name, diagnosis, symptoms, or anything discussed in session, it's Protected Health Information. The service that receives and processes it is now a Business Associate under HIPAA, which means you need a signed Business Associate Agreement with them, documented risk analysis, vendor review, and breach-notification exposure if their systems are ever compromised.
Some cloud services will sign BAAs. Many won't — and the ones that won't are often the fastest, most convenient options. Therapists I've talked to about this describe the same pattern: they try a consumer voice-to-text tool, realize partway in that they're dictating client details into someone's cloud, and then feel stuck between “this saved me an hour today” and “this is probably not OK.”
The honest version: A lot of therapists in solo practice are using voice-to-text tools right now that are not HIPAA-ready, because the time savings are real and the compliance risk feels abstract. It's abstract until it isn't. The architecture below makes the question moot.
The Local-Processing Alternative
A different architecture exists on modern Macs: dictation where the audio never leaves your machine. Apple Silicon can run OpenAI's Whisper speech recognition models locally using the Neural Engine. The audio is processed in memory on your MacBook, transcribed to text, and the audio itself is discarded. No server involved.
Under HIPAA, a software tool that never receives, transmits, or stores PHI is not a Business Associate. There's nothing for a BAA to govern because there's no PHI being shared with the vendor. This is the same architecture we laid out for physicians in the HIPAA-compliant dictation piece; it applies cleanly to therapists too.
EmberType is built on this model. From the design-decision side: one reason we went this route is that the first question a therapist asked us during beta testing was “where does the audio go?” When the answer is “nowhere,” the rest of the conversation is just about speed and accuracy. That's a better conversation to have.
Progress Notes vs. Psychotherapy Notes — The Part Most Articles Skip
HIPAA treats two kinds of therapist documentation differently, and the difference affects how you should think about dictation.
Progress notes (sometimes called session notes, clinical notes, or chart notes) are part of the client's medical record. They document what you did in session at a level appropriate for continuity of care, billing, and disclosure. These are subject to standard HIPAA disclosure rules, and clients generally have access to them.
Psychotherapy notes (sometimes called process notes) are the clinician's private working notes — your observations, hypotheses, analysis, and working therapeutic thinking. HHS guidance gives psychotherapy notes special protection — they are kept separate from the rest of the medical record and are generally not released even when the rest of the chart is.
For dictation, the implications are practical:
- Progress notes go into your EHR or practice management system as structured chart entries. They're concise, template-driven, and benefit enormously from voice dictation speed.
- Psychotherapy notes go somewhere separate — a locked file, an encrypted note app, a dedicated section of the EHR that segregates process notes. They may be more exploratory and loosely structured.
- A local-processing dictation tool works for both, because the audio never leaves your machine either way. A cloud dictation tool complicates both, because you've now routed two categories of sensitive material through a third party.
A Template-Driven SOAP Workflow That Takes 4 Minutes
Most therapists who dictate effectively use a template-driven approach rather than trying to freestyle every note. Here's the workflow a few clinicians we've talked to have converged on — SOAP format, voice-first.
Step 1 — Between clients, speak the scaffolding
In the two or three minutes between clients, dictate a skeleton like this into your EHR's note field:
“Subjective. Client reports... Objective. Presentation was... Assessment. Mood and affect... Plan. Continue current treatment plan with focus on...”
The section headers anchor your thinking. You don't need to finish the sentences in this pass — you're just laying down the frame while the session is still fresh.
Step 2 — Fill each section in a single dictation pass
Go back through the skeleton section-by-section and speak the content. Subjective first (what the client said about their experience). Objective next (your clinical observations). Assessment (your clinical impression). Plan (what you're doing next session).
Speak in complete sentences. Pause for a second between sentences to let the punctuation land cleanly. Don't try to edit while you dictate — get the content down, then review.
Step 3 — A 60-second clean pass
Scan the note. Fix any transcription errors (clinical terms that got mis-recognized, names that got misheard). Add the billing code, CPT, or whatever structured fields your EHR requires. Save.
A note done this way runs 3–5 minutes for a straightforward session, maybe 6 minutes for a complex one. That's a meaningful difference from the 10–12 minutes a typed note takes.
Custom Vocabulary: The Make-or-Break Setting
Whisper, which is the underlying speech model for EmberType and most modern Mac dictation tools, is a general-purpose model. It has been trained on enormous amounts of spoken English but it hasn't been specifically tuned on your clinical vocabulary.
That means out of the box it will sometimes misrecognize specific medications, clinical scales, and proper names. The solution is a custom dictionary: a list of terms that the dictation tool prioritizes. For therapists, the high-leverage entries are:
- DSM codes you use often (F41.1, F33.1, F43.10, etc.) — these often get transcribed as natural numbers without the format
- Medication names your clients are prescribed — sertraline, lamotrigine, bupropion, quetiapine, etc.
- Clinical instruments and scales — PHQ-9, GAD-7, PCL-5, ACE, WHODAS
- Treatment modality shorthand — CBT, DBT, EMDR, IFS, ACT, MI
- Client first names you see frequently — if you have three clients named “Raya” you don't want “Raya” transcribed as “Raya, Rayah, and Ria” across three different notes
Ten minutes spent populating a custom dictionary will save you an hour a month of correction time. It's the single highest-leverage configuration step for clinical dictation.
EHR Compatibility: Where Does the Note Actually Go?
System-wide voice dictation on Mac types into whatever text field has focus. That means any EHR with Mac-compatible note fields works — no EHR-specific plugin required. A quick tour of the common options:
- SimplePractice — web-based. The progress-note fields accept system-wide dictation directly. Many therapists dictate the SOAP body into the Progress Note field.
- TherapyNotes — web-based. Same story: any text field accepts dictated input.
- TheraNest — web-based. Progress notes, treatment plans, assessments all accept dictated input.
- Valant — web-based. A bit heavier UI; the narrative note fields still accept voice input.
- ICANotes, Kareo, Practice Fusion — all web-based, all work.
If your EHR has a text field you can type into, you can dictate into it. The dictation tool doesn't need to know anything about the EHR.
What About Summarization and AI Assistants?
A growing category of clinical AI tools — DeepScribe, Abridge, Augmedix, and similar — records sessions and generates structured notes automatically. These are legitimate tools that offer signed BAAs and are used extensively, especially in larger practices.
The tradeoffs:
- They record the session itself, not just your post-session dictation. That changes client consent, and some clients object strongly to having sessions recorded even with BAAs.
- Session audio is transmitted to and retained by the vendor, requiring BAA-covered vendor management.
- The generated notes still need clinical review. AI-drafted notes have a “feels right, probably wrong” failure mode that requires careful reading, which eats back some of the time savings.
Voice dictation is a different tool for a different preference. If you want your notes to reflect your own clinical thinking in your own words — and you want to avoid recording the session itself — dictating a post-session note yourself is usually the right shape for that.
The Setup: 15 Minutes to Your First Dictated Note
- Download EmberType. Free 7-day trial. Apple Silicon Mac, macOS 14+.
- Pick a Whisper model. For clinical dictation, Large v3 handles medical terminology best. Medium is faster and fine for casual use.
- Set a global shortcut.
Option+Spaceis common. Pick something you won't accidentally hit in the middle of an EHR form. - Add your custom vocabulary — medications, DSM codes, clinical scales, frequent client names.
- Test in a non-PHI context first. Dictate a generic paragraph of clinical-sounding text and check accuracy. Adjust your speaking pace if needed.
- Dictate your first real progress note between sessions tomorrow. Most therapists are at full speed within a week.
Bottom Line
The reason therapists dread the last hour of the day isn't that the notes are intellectually hard. It's that typing is slow, the notes are many, and the combination is exhausting. Voice dictation fixes the typing part. Local processing fixes the privacy part. Put both together and you get a workflow that shortens your day without adding vendor exposure to your practice.
It is not a panacea — you still have to think clearly, document accurately, and review what you dictated. It just removes the keyboard bottleneck. For most therapists who try it, that's enough.
Free Mac Dictation Tips
Get tips on voice-to-text, dictation workflows, and productivity. No spam.
Unsubscribe anytime. We never share your email.
You're in! Check your inbox.
Dictate Session Notes on Your Mac
Local Whisper AI. No audio leaves your machine. 7-day free trial.
Download EmberType FreemacOS 14+ required. Apple Silicon only. $29 after trial.
