Use voice input
import DataBoundary from ‘../../../components/DataBoundary.astro’;
Outcome
Section titled “Outcome”Your spoken question is locally transcribed and inserted into the chat input for you to review before sending.
Prerequisites
Section titled “Prerequisites”- Microphone permission for Synthezia.
- An installed local Whisper model.
- An open session chat or Global Chat input.
- Select the microphone button beside the chat input.
- Speak your question. The control shows that recording is active.
- Select it again to stop and transcribe your speech.
- Review and edit the inserted text.
- Send it only after the text accurately reflects your question.
Expected result
Section titled “Expected result”Synthezia records microphone-only voice input, transcribes it locally, and places the result in the chat input. It does not capture system audio for this workflow.
Data boundary
Section titled “Data boundary”The voice-input transcription is local. Sending the resulting chat question follows the active Local Mode or External API chat boundary.
Common problems
Section titled “Common problems”- The microphone button is unavailable: wait for an in-progress chat or voice transcription to finish.
- The input is blank or inaccurate: check microphone permission and use a clearer, shorter utterance before sending.
- A model dialog appears: download the requested local Whisper model, then retry.