Back to home

Setup guide

Start talking to Hermes with Hey Stentor.

Use this guide to install the desktop app, connect Hermes, configure voice providers, and tune hands-free conversation behavior.

Download the macOS app Download for Ubuntu/Debian Download the AppImage
1

Install the app

Download the Apple silicon macOS DMG or the Linux x64 DEB package. On macOS, move Hey Stentor to Applications and allow the first launch in System Settings if needed. On Ubuntu or Debian, use the DEB package. Use the AppImage when you want a portable executable instead.

2

Grant microphone access

Open Settings > Device, grant microphone access, then use the voice calibration tools to confirm that input level and speech-end detection respond to your normal speaking volume.

3

Connect Hermes

Keep Hermes running locally or through your configured callback route. In Settings > Hermes, confirm the endpoint and shared secret before sending spoken turns.

4

Choose STT

In Settings > Voice, use ElevenLabs as the recommended realtime STT provider. OpenAI and Deepgram remain available when you want to compare provider behavior.

5

Choose TTS

Select ElevenLabs or Cartesia as the speech output provider, add the API key and voice ID, then run a probe before starting a full voice session. OpenAI TTS is planned.

6

Speak and interrupt

Use Speak mode for hands-free capture. Talk naturally, interrupt the assistant when needed, and adjust speech-end sensitivity if capture ends too early or waits too long.

Realtime STT

Shows speech as text while you are still talking.

Role
Captures microphone audio and streams partial text into the conversation before the final transcript is complete.
Where to configure
Use Settings > Voice to select ElevenLabs for the recommended default, or OpenAI if you want its realtime delay and chunk controls.
When to tune
ElevenLabs does not need separate latency or accuracy tuning in the app today. Tune OpenAI only if live text appears too slowly, misses early words, or needs more context.

Troubleshooting

Quick fixes when voice does not feel right.

No microphone input

Recheck macOS microphone permission, then use Settings > Device to confirm the level meter moves.

Speech ends too early

Lower speech-end sensitivity or increase speech-end delay in Settings > Voice.

No spoken reply

Verify the selected TTS provider has a valid API key and voice ID, then run the provider probe.