Voice Input

Genfeed prompt bars can record microphone audio in supported web browsers and send the recording to the authenticated API for transcription.

Requirements

• a browser with MediaRecorder and navigator.mediaDevices.getUserMedia;
• microphone permission for the Genfeed web origin;
• an authenticated API session; and
• a configured transcription path on the API.

Browser and operating-system support varies. The UI hides or disables voice recording when the required browser APIs are unavailable.

Use voice input

1. Open a prompt bar that exposes the microphone control.
2. Allow microphone access when the browser asks.
3. Start recording and speak the prompt.
4. Stop recording and wait for transcription.
5. Review the inserted text before generating.

The recorder captures WebM/Opus audio when the browser supports it. Recordings larger than 25 MB are rejected by the client. Transcription can consume credits, as shown by the product notification after a successful request.

Privacy

This is backend transcription, not guaranteed on-device speech recognition. Audio leaves the browser and is handled by the configured Genfeed API and its transcription provider. Self-hosters should review that provider’s data terms.

Troubleshooting

• If permission is denied, allow microphone access in the browser/site settings.
• If no microphone is found, connect one and reload the page.
• If recording is unsupported, use a current browser with MediaRecorder support.
• If transcription fails, confirm the API session and transcription-provider configuration.

The repository also contains a lower-level Web Speech API hook for feature code, but these docs do not claim a mobile-app voice workflow, continuous listening, or a plan-specific local-processing mode.