Is Aqua Voice on-device?
Short answer: no. Aqua Voice is a cloud dictation service. Audio is captured on the device and transmitted to Aqua’s servers for transcription; the speech model runs in a data centre. That is the architectural choice the product is built around, not a configurable behaviour. For users who need on-device dictation specifically, Aqua Voice is not the shape of product to procure.
What "cloud dictation" actually means
Cloud dictation products run the speech model in a data centre. The client app records audio, sends it to the service over the network, receives text back. Aqua Voice fits this shape. The architectural advantage is that the cloud can use larger models, update them server-side and offer broader language coverage; the operational consequence is that audio leaves the device and lives, however briefly, on someone else’s infrastructure.
For dictation that does not involve sensitive content, this trade is broadly fine. The vendor’s privacy policy describes how audio is processed and retained, and modern cloud transcription services typically advertise reasonable defaults. Where it matters is when the content itself has obligations attached.
Where the cloud architecture changes the conversation
For Protected Health Information under HIPAA, the cloud transcription vendor would need a Business Associate Agreement in place with the covered entity. For UK GDPR, voice data that can identify a speaker is biometric data when used for identification, and any third-party processing needs a lawful basis and a written processor agreement. For privileged matter, introducing a third party into the audio chain is a deliberate decision that needs to fit the firm’s position on cloud processing of client work.
None of these are insurmountable, and many cloud dictation products serve regulated industries under the appropriate contracts. They are the operational cost of choosing the cloud architecture over the on-device one.
The fuller framing is in HIPAA and dictation: architectural vs contractual privacy.
The on-device alternative
Parakeety is built around the opposite architectural choice. Parakeet TDT v3 runs on the Apple Neural Engine on your Mac. Audio is captured to memory, transcribed locally and discarded. No data centre is involved. The only outbound traffic the app ever produces is a one-time speech-model download and periodic license checks; neither carries audio.
For users who need on-device dictation on a Mac, that is the architectural answer. The trade is the standard cloud-vs-local one: Parakeety is Mac-only and covers 25 European languages, against the broader language coverage and feature surface a cloud product can offer.
FAQ
- Is Aqua Voice on-device?
- No. Aqua Voice is a cloud dictation service. Audio is captured on the device and transmitted to Aqua’s servers for transcription. The speech model runs in a data centre, not on your Mac. That is the architectural choice the product is built around.
- What does cloud architecture imply for privacy?
- Audio leaves the device. That introduces a third party that processes your audio. For most users this is a normal trade for a cloud product; for Protected Health Information, privileged matter or NDA-bound content, it is a third-party processor that needs to be evaluated under the relevant regime (HIPAA, GDPR, professional codes of conduct).
- Can I configure Aqua Voice to run locally?
- Not at this time. The product is built around cloud-hosted models and the architecture is not designed for on-device transcription. A future on-device option is not commonly advertised; for now, "Aqua Voice" and "on-device" are mutually exclusive.
- What is the on-device alternative on Mac?
- Parakeety runs Parakeet TDT v3 on the Apple Neural Engine. Audio is captured to memory, transcribed locally, pasted at the cursor and discarded. After the one-time speech-model download, no audio leaves the Mac. For users who need an on-device dictation tool, that is the architectural answer.
Try Parakeety
Parakeety is a Mac menu-bar app. Hold the section key, talk, release; your words paste at the cursor in whichever app you were typing into. Audio never leaves the machine. There is a free 7-day trial with no card required. After that it is $30 once.