Skip to content

Voice Input

The mobile app supports voice input through a tap-to-record button in the chat input bar. Tap to start recording, tap again to stop and send.

  1. Tap the microphone icon in the chat input bar
  2. A recording indicator with an elapsed-time counter replaces the input, and a haptic tap confirms recording started
  3. Speak your message naturally
  4. Tap the indicator to stop — the audio is transcribed and sent as a chat message
  5. The agent responds with a streaming text reply

Transcription happens on your daemon, not on the phone. The app uploads the recording to your daemon, which transcribes it with whichever speech-to-text provider you’ve configured there:

ProviderWhere it runs
Local WhisperOn your daemon’s machine — audio never leaves your hardware
OpenAI WhisperOpenAI’s API, using your key
ElevenLabsElevenLabs’ API, using your key

No Snippbot servers are involved. Configure providers and API keys in the desktop UI.

Settings → Voice shows the speech-to-text and text-to-speech providers your daemon has, and whether each one is configured. The screen is read-only — provider setup, keys, and voice selection are managed on desktop.

  1. Speak clearly and at a natural pace
  2. Minimize background noise — find a quiet environment
  3. Hold the phone 6–12 inches from your mouth
  4. Use short phrases for best accuracy
  • Voice input requires microphone permission (requested the first time you record)
  • Recording requires a physical device — not available in emulators
  • Your daemon needs a speech-to-text provider configured; transcription accuracy depends on that provider
  • Long recordings take longer to upload and transcribe, especially over a remote tunnel
  • Spoken (text-to-speech) agent replies are not available on mobile yet
  1. Check microphone permission in device Settings > Apps > Snippbot
  2. Ensure no other app is using the microphone
  3. Try restarting the app
  1. Open Settings → Voice and check that a speech-to-text provider shows as configured
  2. Check your daemon connection in Settings → Connection
  1. Speak more slowly and clearly
  2. Reduce background noise
  3. Try shorter phrases
  4. Try a different speech-to-text provider on your daemon