openclaw - 💡(How to fix) Fix Feature: Dual voice input buttons (auto-send + continuous mode) [1 participants]

Feature Request: Dual Voice Input Buttons for Control UI

Current Behavior

The Control UI has a single microphone button for speech input. After speaking, the recognized text appears in the input field, but the user must manually press Enter to send the message. The mic also stays active until manually stopped.

This creates friction:

• Extra step (press Enter) after every voice message
• Mic doesn't auto-stop, so you have to click again to turn it off
• No way to have a "conversation mode" where you speak multiple messages in sequence

Proposed Behavior

Replace the single mic button with two side-by-side mic buttons:

Left button — "Voice Auto-Send" (one-shot)

• Click → mic activates
• Speak → pause detected → text is sent automatically
• Mic turns off after sending
• Ideal for quick single messages

Right button — "Voice Continuous" (conversation mode)

• Click → mic activates and stays on
• After each speech pause, text is sent automatically
• Mic stays active — just keep talking for the next message
• Click again to deactivate
• Ideal for longer back-and-forth conversations

Both modes show a live interim transcript above the input field while recording.

Implementation Notes

This was prototyped as a patch to the built Control UI (dist/control-ui/assets/index-*.js) with these changes:

1. Added sttRecordingCont state variable alongside existing sttRecording
2. Updated interim transcript display to check both recording states
3. Split the single mic button into two:
   • Auto-send button: sets MC.continuous = false, calls e.onSend() + PC() on final transcript
   • Continuous button: on onEnd, restarts recognition via window.__sttCont.start() for persistent listening
4. Both buttons use the existing NC() / PC() / jC() infrastructure — no new APIs needed

The patch works reliably on Chrome and Edge (Web Speech API). After each OpenClaw update, the patch must be re-applied since the built JS filename hash changes.

Why This Matters

Voice input is one of the most natural ways to interact with an AI assistant. The current UX (speak → type appears → press Enter → stop mic) feels like fighting the tool. Auto-send on pause makes voice feel as fluid as talking to a person. The continuous mode is especially valuable for extended conversations where you don't want to re-activate the mic every 30 seconds.

Environment

• OpenClaw 2026.4.x
• Chrome / Edge (Web Speech API)
• Tested with the existing sttRecording / NC / PC speech recognition infrastructure