Medical Transcription for Dictation and Interviews

Author: Hushscript | Published: July 17, 2026 | Last reviewed: July 21, 2026

Medical transcription covers two different jobs that happen to share a name: a clinician dictating notes alone into a recorder, and a researcher or clinician recording a conversation with a patient or colleague. Hushscript handles both the way it handles any recording. You upload it and get a transcript back, with one specialized mode for clinical language, and one boundary worth knowing before you rely on it for something sensitive.

What medical mode actually does

Turn on medical mode before you start a job that’s dictation or a clinical interview, and the model behind the transcription switches to one built for clinical language: drug names, anatomical terms, dosages, and the shorthand a clinician says without thinking twice about it. It runs as a specialized recognition pass instead of the general-purpose model that handles everything else, which is why it’s an option you switch on per file rather than a setting that changes every transcript by default. That specialization has a cost. Medical mode doubles the billed minutes for the file it’s turned on for, and only for that file: a 20-minute dictation billed with medical mode on draws 40 minutes from your balance, while everything else you transcribe without it stays at the normal rate. It draws from the same balance as any other job, and that balance starts with 30 free minutes once you verify a payment method. Adding a card places a $1 hold that’s authorized and then released without a charge; pay another way and the 30 minutes arrive with your first purchase. After that, it’s pay-as-you-go.

Saved dictionaries for clinical vocabulary

Medical mode changes which model listens to the recording. A saved dictionary changes what it’s listening for. Build one from the drug names, procedure terms, and abbreviations specific to your practice or study, import it in bulk from a CSV or TSV file rather than typing each term by hand, and organize entries into categories with spelling variants attached to each canonical term. That helps for a drug name that gets said differently than it’s spelled, or a term with more than one accepted abbreviation. Mark the dictionary as your default and it’s preselected automatically on every upload afterward, so a standing clinical glossary doesn’t need reselecting file by file. It combines freely with a one-off keyterm list for a term specific to a single recording, and with medical mode itself; the two aren’t a choice between one or the other.

Where private mode and medical mode don’t overlap

Private mode exists for the recording that shouldn’t leave a saved transcript behind at all. You get a password-protected archive instead, with speech recognition only and nothing else running on the result. For a sensitive clinical recording, that’s often exactly the setting you want. But medical mode isn’t available inside private mode: the specialized clinical model is one of two things private mode’s speech-recognition-only path skips, translation being the other. If a file needs both clinical vocabulary handling and the shortest possible trail afterward, know that trade-off going in. You get one or the other on that upload, not both at once.

Retention for the recordings that stay

Outside private mode, a normal transcript follows a retention window you set. Keep everything until you delete it yourself, or auto-remove it after 7, 30, 90, or 365 days, with any single transcript able to carry its own override. The audio itself doesn’t wait on any of that: it’s deleted the moment the transcript is ready, on every job, clinical or not. Deleting one transcript through its dedicated flow returns a receipt recording what was removed and when; scheduled and bulk deletion do not return one yet. That record is one you can keep for your own files, not an independent audit of anything beyond Hushscript’s own side.

A worked example

A clinical researcher records a 25-minute structured interview with a study participant, covering medication history and symptom timeline, dense with drug names and dosage terms a general model would likely mishear. They turn on medical mode before uploading and load their study’s saved dictionary, built once from a spreadsheet of the drug and condition names the interview protocol covers. The transcript comes back with speaker labels separating interviewer from participant, the drug names landing correctly on the first pass, and medical mode’s cost reflected as 50 billed minutes for the 25-minute file. They skim it against the recording once, fix the two terms the dictionary didn’t cover with a find-and-replace that catches every instance, and export it for the study file.

What this isn’t

Everything above (a specialized recognition model, saved dictionaries, retention controls, deletion receipts) is a description of what Hushscript does, not a claim about what any specific organization’s rules require. Those requirements come from sources such as 45 CFR Part 164, interpreted in the context of your organization; NIST SP 800-66 Rev. 2 is implementation guidance, not a product badge. Hushscript does not claim regulatory status. Treat these as capabilities, not guarantees, and have the responsible person in your organization assess them before regulated use.

Where to start

Medical mode, saved dictionaries, and the rest of the upload options live on audio to text alongside every other per-file setting. For the recordings that need to skip a saved transcript entirely, private transcription covers private mode and the retention controls in full. If building a standing clinical dictionary is new territory, teach the transcriber your vocabulary goes deeper on the mechanics, and for the private-mode trade-offs mentioned above, private mode: what it does and skips has the complete picture.