[Feature Request] Push-to-Talk Hotkey #11
Description
Push-to-Talk Hotkey
Description
Use a hotkey to let the AI listen only. Release the hotkey to send the request.
Details
- Press and hold the hotkey to open the mic and start listening. When the user releases the hotkey, the mic closes, and the request is sent. This flow should bypass VAD.
- The hotkey works in both the Web app and the Electron app. In Electron, it remains usable even if the Open-LLM-VTuber client is minimized.
- If the mic is open when the hotkey is pressed, the mic should remain open after releasing the hotkey. Note that we should bypass the VAD whenever we use the hotkey. We should not send the request before the user releases the hotkey.
- If the mic is not open when the hotkey is pressed, the mic should be turned off after releasing the hotkey.
- If possible, the mic button should be in a different color (or just keep changing color) when the hotkey is pressed, so the user knows that the mic is in a weird state.
- The user should be able to customize the hotkey.
Why
This encourages users to keep Open-LLM-VTuber running in the background and talk to it on demand—without misfires or unexpected responses.
- Ensures the AI responds only when the user intends it to.
- Encourages the user to run Open-LLM-VTuber in the background all the time.
- Make it super easy to talk to Open-LLM-VTuber.
- Provides an alternative to VAD, which can be inaccurate in noisy environments and may cut off longer or stuttered speech.
Will I be able to work on this issue?
No. Codex failed me on this feature. I have no idea how to make this happen.