Your audio never leaves your device
Transcription for therapists, with sessions kept on your device
Turn recorded sessions into text for your notes without uploading the audio anywhere. The transcription happens in your browser, so a sensitive recording never leaves your control.
Takes a little extra time, and your audio still never leaves your device. Labels go by how voices sound, so treat them as a best effort: one person can be split across two or three labels, similar voices can be merged, and someone joining partway through may be folded into a speaker already talking. Giving a number usually helps. Check them before relying on them.
Specifying this number improves accuracy. Count people who really take part, whenever they join. Someone who only says a word or two is better left out.
Timestamps each word instead of each phrase, so short interjections like "Yeah" go to the right person. On a busy four-person recording this corrected about 1 word in 5; it changes little when people talk in long turns. Adds roughly 40% to the time on a computer, and almost nothing on a phone.
Phone audio loses the high frequencies that separate consonants, so they are put back before transcribing. Only phone-quality recordings are treated, and you are told when it happens. Also even out volume is for one case: someone far from the microphone and much quieter throughout. It cost accuracy on other recordings, so use it only if a quiet person came out badly.
Worth filling in. The speech model knows ordinary English, not your subject, so a word it has never heard becomes whatever sounds closest, and it usually gets that word wrong every single time it comes up. A sentence is enough, and the names and jargon in it are picked out and watched for. It cannot guess a word you do not mention, so name the people and the terms that matter.
Free, no signup. The first time you use it, it takes a minute or two to set up. After that setup is quick. Works with MP3, WAV, M4A, OGG, FLAC and AMR, and long recordings are handled in parts.
Confidentiality without the upload
Session recordings are among the most sensitive audio anyone handles. They contain identifiable detail about a person who disclosed it under an explicit promise of confidentiality, often including third parties who never consented to anything. Uploading that to a transcription company means extending your confidentiality obligation to a vendor your client never chose.
MeetingScribe removes that step entirely. The model runs in your browser, the session is transcribed on your machine, and the text is yours alone. Nothing is stored on our side because nothing reaches our side. If you want to confirm that rather than accept it, how it works shows you how to watch the network yourself and how to run the whole thing with your internet disconnected.
What clinicians actually use a transcript for
A verbatim transcript is not a progress note, and treating it as one creates more problems than it solves. The uses that hold up in practice are narrower:
- Writing your own notes accurately. Recall degrades fast, and a transcript lets you write a note from what was said rather than what you remember an hour later.
- Supervision and consultation. Bringing an exact exchange to supervision is more useful than paraphrasing it, assuming consent covers that use.
- Reflective practice. Reading your own interventions in text is uncomfortable and unusually instructive. Patterns in how you respond are visible on the page in a way they never are in the room.
- Training and skills development. Reviewing a session against a model or protocol is far easier with searchable text.
- Accuracy on specifics. Medication names, dates, and exact wording of a disclosure, where getting it wrong in the record matters.
A workflow that ends with the recording deleted
The transcript is usually the artifact you want. The audio is usually the artifact you want gone. A sequence that respects both:
- Record with documented consent covering what the recording is for, who will hear it, and when it will be destroyed.
- Transcribe on the same device that holds the recording. Moving the file to another machine or a shared drive to process it creates exactly the extra copy you were avoiding.
- Split long sessions. A fifty minute session is at the edge of what a browser tab handles comfortably. Two parts named in order, selected together, join into one transcript reliably.
- Correct the transcript in the page. Clinical terms, medication names, and proper nouns are what speech models get wrong most. Fix them before you save anything.
- Consider pseudonymizing as you go. Replacing names with initials or role labels at this stage costs a minute and materially reduces what a stray file could reveal.
- Write your note, then destroy the recording according to your retention policy. The transcript inherits the sensitivity of the audio, so store it wherever your clinical records already live, not in a downloads folder.
A direct word about AI summaries
AI scribes for clinical notes are everywhere right now, and most of them work by sending the session to a server. This tool deliberately does not do that. After transcribing, it can copy a prompt to your clipboard so you can paste it into an AI chat yourself. Nothing is transmitted by the copy.
Pasting session content into a consumer AI service, though, is a genuine disclosure of client information to a third party, whatever the interface makes it feel like. For clinical material that is a decision to make deliberately, with your regulator's guidance in mind and quite possibly with your client's knowledge. If in doubt, use the transcript and write the note yourself. That path never involves anyone but you.
Getting a clean recording in a therapy room
Consulting rooms are acoustically forgiving, which helps, but a few habits noticeably improve the transcript:
- Place the recorder between you and the client rather than beside one of you, so both voices arrive at similar volume. Uneven levels are the main cause of confused speaker labels.
- Choose the small model for quiet speakers, distress, or accented speech. The extra download is worth it when the material is hard to hear.
- Set the language explicitly rather than relying on auto-detection, especially if the first minute is quiet.
- Expect overlapping speech to be the weak point. In couples and family work, plan to correct speaker labels rather than trusting them.
Your obligations stay yours
Different regions, licensing boards, and professional bodies have their own rules for recording, consent, what belongs in a record, and how long material may be kept and where. MeetingScribe is a private way to convert audio you are permitted to use into text. It does not replace your judgment about whether to record, does not claim any certification, and is not clinical or legal advice. Follow the standards that apply to your practice.
Last reviewed: August 2026
Frequently asked questions
- Is this compliant with health privacy rules?
- We make no compliance claims and provide no business associate agreement. Keeping audio on your device is why the design suits confidential clinical work, but meeting your own privacy, consent, and record-keeping obligations remains your responsibility.
- Do I need client consent to record?
- Yes. Follow the informed-consent and record-keeping rules that apply to your licence, your professional body, and your jurisdiction. This tool transcribes audio you already have the right to use; it does not advise on consent.
- Should the transcript go into the client record?
- That is a clinical and regulatory judgment, not a software one. In some frameworks a verbatim transcript is treated differently from a progress note, and creating one may change what is disclosable. Decide deliberately, and check your regulator's guidance rather than defaulting to saving everything.
- Will it separate me from the client?
- Usually yes. Speaker separation is most accurate with one to three voices, which fits individual sessions well. Couples and family sessions with overlapping speech are harder, and you can correct the labels yourself.