Where transcription, the gateway and account data each happen

Different features follow different paths. To judge a privacy boundary, first ask what the feature does right now, rather than summing it up in a slogan. The paths are explained separately below.

1. Transcription path

Transcription runs on your device by default and the audio stays on your machine; when you choose the cloud fast or standard mode, audio is transcribed in the cloud through the gateway and the content is not stored. This describes only the speech-to-text step; answer generation and the capabilities after it follow their own paths.

2. Cloud answers

When an answer has to be generated, the current question and the context you have enabled are handled through the gateway to produce something you can say as written. Usage is deducted from your plan quota; see the pricing page for the quota rules.

3. Screen context

Screen reading can be turned off on its own at any time. When it is on, the current material takes part in the answer background; when it is off, this source is not used as context. Visibility while sharing should still be verified against your actual system and meeting app.

4. Account, device and controlled download

The customer center handles sign-in, balance, plan, device binding and download. After authentication the installer gets a signed link valid for 5 minutes, with a user watermark where it can be applied.