Last updated October 19, 2026. Documentation takes a large share of a clinician's day, and in a small practice there is rarely a department to absorb it. Medical transcription is the oldest answer to that problem: you speak, someone or something turns it into a written note. This explainer covers how the different approaches work today, what they cost in time and attention, which privacy questions to ask any vendor, and how a small practice can set up a routine that holds up.
What medical transcription means today
Traditionally, medical transcription meant a clinician recording dictation on a handheld recorder or a phone line, and a trained medical transcriptionist typing it into a formatted report: a consultation letter, an operative note, a discharge summary. The transcriptionist knew the terminology, the report templates and the habits of each doctor, and returned the document for signature.
That model still exists, but most practices now use one of three patterns, often mixed.
- Human transcription services, where audio is sent to trained transcriptionists who return typed reports.
- Speech recognition with self editing, where the clinician dictates and the text appears on screen, then the clinician corrects it.
- AI transcription with formatting, where software turns dictation or a recorded conversation into a structured draft that the clinician reviews and signs.
None of these removes the clinician's responsibility for the content. The difference is how much of the typing, formatting and checking moves off the clinician's desk.
Human medical transcription services
Human services remain valuable for complex reports and for clinicians who prefer not to edit at all. Turnaround is measured in hours or days rather than seconds, pricing is usually per line or per minute of audio, and quality depends on the individual transcriptionist and the clarity of the recording. The practice also has to manage how audio is sent, who has access and how long the service keeps files.
Speech recognition and AI transcription
Software turns speech into text immediately. The trade off is that the clinician, or a staff member, reviews and corrects every note. Modern AI tools go further than word by word recognition: they add punctuation, separate paragraphs, recognize a requested structure such as a referral letter, and learn custom terms. That shifts the work from typing to reading, which for most clinicians is the faster of the two.
Dictation versus recorded encounters
There is an important distinction between dictating a note after the visit and recording the visit itself. Dictation captures only the clinician's voice, summarizing what matters. Recording an encounter captures the patient's voice too, which raises consent questions and produces much more raw text to review. Many small practices find that structured dictation after each visit gives most of the time savings with far fewer complications.
Medical terminology and accuracy
General speech recognition struggles with drug names, dosages, anatomical terms and eponyms, which are exactly the words where an error matters. Any tool you consider should let you add your own terms, including the brand names you prescribe often, colleagues' surnames for referrals and the abbreviations your practice actually uses.
Stenograph includes professional vocabularies for medicine and lets each user add custom terms (500 on Solo, 2,000 on Pro, shared dictionaries on Team). Even so, numbers deserve special care. Read back every dose, frequency and laterality before signing. A transcript that says "15 mg" when you said "50 mg" looks perfectly plausible on screen.
Privacy questions to ask any transcription vendor
Patient information is among the most sensitive data a practice handles, and the health privacy rules that apply to you depend on your country, your state and your role. Before adopting any tool, ask the vendor these questions in writing and compare the answers with your organization's policy.
- Is the audio stored after processing, and if so, where and for how long?
- Is the resulting text stored, can it be switched off, and can it be deleted on request?
- Is any customer content used to train AI models?
- Is data encrypted in transit and at rest?
- Which categories of subprocessors receive the data, and in which regions?
- What certifications or contractual commitments does the vendor offer, and do they meet what your practice requires?
How Stenograph handles audio and text
Stenograph does not claim a health data certification. Here is exactly what happens. Audio is processed and then deleted, it is not stored. Text history is optional: you can turn it off so nothing is kept after insertion, and you can delete existing history at any time. Content is encrypted in transit and at rest, and it is never used to train AI models. Clinicians should follow their organization's policy on patient identifiers, which for many practices means dictating without names or record numbers and adding those inside the record system. The dictation privacy page has the full description.
A practical medical transcription routine for a small practice
A routine matters more than the tool. This is a pattern that works for many solo clinicians and small teams.
- Dictate immediately after each visit, while details are fresh, in a consistent order such as reason for visit, findings, assessment and plan.
- Use a spoken template phrase, for example "referral letter to cardiology", so the output follows the right structure.
- Dictate into the place the note will live, at the cursor, rather than into a separate app you copy from.
- Review every note before signing, with extra attention to medications, doses, numbers and left or right.
- Keep a shared list of terms and colleague names so the vocabulary improves for everyone in the practice.
- Decide once, in writing, what identifiers may be dictated and where text history is kept.
With Stenograph the dictation step is one hotkey on Mac or Windows, or one tap on the iPhone and Android keyboard. The microphone is on only while you speak, and formatted text lands in the app in front of you, whether that is your record system in a browser, a word processor or an email. The details for clinicians are on medical dictation software.
Transcribing recorded dictation and meetings
Some clinicians prefer to dictate into a recorder during the day and process the files later. File transcription handles that pattern: upload the audio, get a transcript with timestamps, and export to Word, PDF or TXT. For practice meetings and case discussions, recording on your own device produces a transcript with speaker labels, a summary and action items. Our page on medical transcription software covers file and meeting workflows for practices.
Common documents in small practice transcription
Most of what a small practice dictates falls into a few document types, and each benefits from a consistent spoken structure.
Visit notes
Speak the note in the order your record system expects. Reason for visit, history, examination findings, assessment, plan. Saying the section names out loud gives the formatter clear boundaries and gives you a predictable layout to review.
Referral letters
A referral letter has a recipient, a reason, relevant history, current medications and the specific question you want answered. Start the dictation with "referral letter to" and the specialty, and finish with the question. The letter comes back with a salutation, paragraphs and a closing, ready for your signature.
Patient letters and instructions
Letters to patients need plain language. Dictate as you would explain it in the room, then ask for a shorter or simpler version if the first draft reads too clinical. Always read medication instructions back against your prescription.
Practice correspondence
Insurance queries, letters to colleagues and internal emails are where dictation saves time without any clinical risk, and they are a good place to build the habit before using it for clinical notes.
Rolling out transcription in a small practice
Start with one clinician and one document type for two weeks. Track how long notes take before and after, and how many corrections each note needs. Build the shared vocabulary from the corrections you make most often. Once the routine works, extend it to other clinicians, and agree on the identifier and text history rules before anyone dictates patient material.
What medical transcription costs a small practice
Costs come in two forms: money and clinician time. Human services charge per line or per minute and add turnaround time. Software charges a subscription and asks for review time. Stenograph costs 49 USD per month on Solo, or 24 USD per month billed yearly, with 1,500 dictation minutes and 300 transcription minutes each month. Pro, at 99 USD monthly or 49 USD per month billed yearly, adds all professional vocabularies and higher limits. Team covers up to five clinicians and staff with shared dictionaries and central billing. Payments are final, and you can cancel any time with access to the end of the paid period. The full comparison is on stenograph pricing.