Back to Reflex Loop

Voice privacy

Recording and transcription

Reflex Loop records your practice audio only after you start a recording. Your recording is sent to Deepgram for transcription only when you manually submit it. Playing, retrying, or uploading a recording does not request a transcription.

Target Answer processing

When this feature is enabled, your speech transcript and timing metadata are sent to OpenAI to select an edit plan for your own recordings. Your audio, account ID, email, and fluency measurements are not sent for this purpose. Response storage is disabled for this request, and the result is not used in fluency measurement.

Live chunk highlighting

During recording, supported browsers may use their built-in Web Speech API to highlight a chunk as you speak. In Chrome and other Chromium browsers, this can send audio to Google's cloud speech-recognition service. This is a separate service from Deepgram and is used only for immediate highlighting.

If live recognition is unavailable

Practice and recording continue normally. Live highlighting may remain off, and its result is never used as canonical chunk evidence or as a formal fluency measurement.