Clinics, privacy & offline

The problem: sensitive conversations and cloud transcription

Clinics, therapists and other practices record dictations, supervision sessions, case reviews or (with consent) client conversations. Typing notes from them takes time, and cloud transcription is the obvious shortcut. But health and personal data usually comes with strict rules about who processes it, where it is stored, and which contracts must be in place before a vendor ever touches it.

For many small practices, the simplest way to reduce that burden is to not send the audio anywhere. Noteo’s Local mode is designed for exactly that: speech-to-text runs on your own computer using the open-source whisper.cpp engine, and the result is saved locally.

Who this workflow is for

Local mode is a fit for privacy-focused teams that record conversations and want text without an upload step:

• Private practices, therapists and counsellors transcribing their own dictations or supervision notes
• Clinic teams documenting internal case discussions
• Researchers handling interview data covered by an ethics protocol
• Legal, HR and social-work teams with similar confidentiality needs

Worked example: end-of-day dictation

A practitioner records short voice memos after each appointment on a work laptop. At the end of the day she opens the Noteo desktop app, which is already set to Local mode, and imports the files one by one. For each file, ffmpeg converts the audio to 16 kHz mono WAV and whisper.cpp transcribes it with the model configured in Offline settings, with the language set to her working language or to auto-detect.

Each transcript is saved to the local library on the laptop. She reviews the text, corrects terminology the model got wrong, and exports a TXT or Markdown file into her practice’s documentation system. The recordings were never uploaded to Noteo.

What Local mode does technically (and what it does not)

It helps to be precise about what “offline” means here, so you can assess it against your own obligations:

• Transcription runs locally: ffmpeg and whisper.cpp are executed on your computer with a model file stored on your disk.
• Local transcripts are saved to a library file in the desktop app’s data folder, not to Noteo’s cloud database.
• During processing, ffmpeg writes a temporary WAV file to the operating system’s temp folder.
• AI summaries and chat are disabled in Local mode. They are Cloud features that you have to switch modes to use.
• Noteo does not encrypt these local files itself, so device security (disk encryption, user accounts, backups) is your responsibility.

Local vs cloud: the trade-offs

Cloud mode adds conveniences such as speaker recognition, AI summaries and chat, and it works from the browser. The cost is that recordings are uploaded and processed on servers, which may require a data processing agreement and a review under your local regulations. Local mode avoids the upload, but transcripts are single-speaker (no diarization), speed depends on your hardware, and you manage the installation yourself.

Compared with cloud-only note takers, where uploading is the only option, Local mode keeps the audio in-house. Compared with typing notes by hand, it gives you a draft to review.

Important caveats

Noteo is a transcription tool, not a medical device or a clinical record system, and we do not claim any healthcare-specific certification. Keeping audio on-device is one technical measure, not compliance on its own. Before using any transcription tool with patient or client data, check your obligations under the rules that apply to you (for example GDPR, national health-data laws or professional codes), get the consent you need, and consult your data protection officer or legal advisor. Always review transcripts before they enter a record, since speech recognition makes mistakes, especially with medication names and specialised terminology.