Offline recording can make clinical audio capture more practical, but it does not establish that speech recognition or note generation happens entirely on a wearable. For Heidi Remote and other scribes, assess the whole journey from capture to the reviewed record.
This article is published by iatroX. Product descriptions were checked on 4 November 2026. It compares documented data paths and questions for evaluation; it does not report a security audit or a trial of the devices.
Three different meanings of offline
A product may record audio while disconnected, transcribe locally or generate a finished note without an internet connection. These are three different capabilities. The first does not establish the other two.
A clinician might record a home visit successfully but still have a pending encounter when they return to the car. Another product might show a transcript before it can produce the structured note. Both workflows can be useful, provided the clinician understands which stage has finished and what remains to do.
The useful questions are concrete: can capture continue, what is stored, what must synchronise and when is the note ready for review? Treat those as separate parts of the evaluation.
What Heidi Remote's current documents establish
Heidi's UK hardware page describes a wearable microphone with offline capture and later synchronisation. It lists £209 before applicable taxes. Its privacy description says audio is encrypted on the device and deleted there after synchronisation. The current Remote help describes the transfer process; synchronisation time depends on the session and connection.
That is a more useful specification than assuming the entire clinical workflow works identically without connectivity. A recording can be complete while synchronisation or note generation remains outstanding.
Do not transfer an app benchmark to a wearable
Argmax's November 2025 case study concerns Heidi's mobile app and on-device speech recognition, including iOS lock-screen operation. Its latency result is not a measured Remote-device upload time or proof that the same recognition model runs on the wearable's own hardware.
When an architectural detail matters to a purchase or data assessment, obtain the specification for the actual device, app version and mode. A supplier's achievement in one component can inform a question about another component; it does not answer it automatically.
What local capture changes
Keeping an unsynchronised recording locally changes where that copy resides and how work can continue during a connection failure. It can reduce dependence on uninterrupted connectivity for capture. It also creates a local item that must remain associated with the correct encounter until processing is complete.
It does not, by itself, eliminate transmission risk: the intended workflow may transfer audio later or synchronise while connected. Nor does it establish that no supplier or processor receives the information. Inspect the later transfer and processing rather than stopping the assessment at the microphone.
Likewise, two products that both use cloud services do not necessarily have identical privacy arrangements. Their contracts, processing locations, access controls, retention and settings can differ. “Cloud” and “on-device” are useful architectural descriptions, but neither is a complete privacy verdict.
Retention needs a data category and a location
Ask separately about raw audio, temporary device backups, transcripts, generated notes, feedback and the final clinical record. They may exist for different purposes and periods. Deleting one copy does not establish deletion of another.
Nabla's current retention help describes configurable patient-data retention, with a standard period of 14 days, and no retained audio. Its separate Offline Mode instructions describe temporary local audio until transcription succeeds. This illustrates why a no-audio-retention slogan should not be expanded into a claim that nothing is ever stored anywhere.
Use the same level of specificity for every supplier. A useful answer identifies the data category, location, purpose, setting and deletion event. A broad claim of zero retention is less informative than a description you can follow through the actual workflow.
Compare the stages rather than awarding privacy scores
| Stage | What to establish | What to observe in a fictional test |
|---|---|---|
| Start capture | How recording is initiated and indicated | Whether the operator can see and stop it |
| Lose connectivity | Which functions continue and what remains locally | Whether the encounter is clearly marked as pending |
| Reconnect | What transfers and how progress is shown | Whether the correct encounter receives the material |
| Generate a note | Where processing occurs and which settings apply | Whether the output is complete enough to review |
| Review and file | Who checks meaning and completes the record | Whether corrections survive in the final destination |
| Close or delete | Which copies are affected by each action | Whether an unsynchronised item remains outstanding |
The table is a proposed evaluation, not a statement that a named product has passed or failed it. A synthetic example lets a team understand the process before involving patient information.
A lost device is an operational problem to plan for
Record who should be contacted, how an account or pairing is secured and what information is needed to identify outstanding sessions. Establish the supplier's recovery and deletion options before a loss occurs.
Encryption is relevant, but its presence alone does not settle the consequences of a particular incident. The organisation needs to know what the device contained, which protection applied and whether another person could access it. Follow the applicable local incident process using those facts.
Do the same for a lost phone or laptop. Physical loss is not a risk exclusive to a dedicated wearable, and a software-only workflow still uses physical equipment.
Patient understanding remains part of the workflow
A small clip-on recorder may be less familiar than a laptop or phone. Explain its purpose and follow the relevant organisation's recording arrangements. Make it clear when recording starts, how a person can raise a concern and what happens if capture is stopped.
An offline mode does not make that conversation unnecessary. Equally, a visible recording light is helpful information rather than proof that every part of the person's concern has been addressed.
Where iatroX fits alongside documentation
iatroX offers a separate reference and learning workflow. A clinician can use free Ask-iatroX to investigate a general clinical question, inspect the linked evidence and follow up a difficult concept through the available learning tools. Paid learning features have their own access terms.
That can complement the documentation route chosen by a practice. A scribe helps produce a record; a referenced learning question helps examine an uncertainty exposed by the work. Modern platforms can combine these functions, so the distinction should be made by feature rather than assuming a competitor only transcribes.
There is no need to send an encounter recording to iatroX to make that learning step useful. Formulate the general question deliberately and review the source's jurisdiction and relevance. The positive case is the connected path from a question to understanding, alongside whichever documentation tool fits the clinical setting.
Choosing a workable setup
Test the whole encounter, including an interrupted connection and a delayed review. Count correction and completion work as well as capture time. A device earns its place when that complete process is easier to manage and the team can tell what remains unfinished.
Offline capture is a useful capability to compare. It should be described precisely, alongside transfer, processing, retention and review, rather than promoted as a universal privacy advantage.