openclaw - 💡(How to fix) Fix [Feature]: Control app option to auto-play assistant TTS audio [2 comments, 2 participants]

Summary

Add a Control app / Control UI setting to automatically play assistant TTS audio attachments when they are generated.

Problem to solve

When messages.tts.auto is set to always, OpenClaw correctly generates MP3 audio for assistant replies, but in the Control app the audio appears as an embedded player that must be manually played.

Current behavior:

• Assistant text reply appears
• TTS MP3 attachment/player appears
• Player initially may show 0 length/no controls while loading
• Audio works after loading, but the user must press play manually

This makes “assistant voice” feel like a manual attachment workflow rather than a conversational spoken assistant experience.

Proposed solution

Add a Control app / Control UI setting such as:

• Auto-play assistant TTS audio
• Possibly scoped per device/session
• Default off, because autoplay with sound can be surprising
• Once enabled by explicit user interaction, automatically play new assistant TTS audio attachments when they arrive

Suggested behavior:

1. User enables auto-play in the Control app/UI.
2. Browser/app satisfies autoplay requirements through that user gesture.
3. Future assistant TTS audio attachments play automatically when ready.
4. If autoplay is blocked by the platform, show a small “Tap to enable voice playback” prompt.

It would also be helpful if the audio player handles loading state more gracefully instead of briefly showing 0 length/no controls.

Alternatives considered

• Manually pressing play on every generated TTS attachment works, but it defeats the purpose of messages.tts.auto = always.
• Using tagged TTS reduces clutter, but does not solve the spoken-assistant use case.
• Platform-level autoplay hacks should be avoided; this should be explicit user opt-in.

Impact

Affected users/systems/channels:

• Control app / Control UI users using messages.tts.auto = always
• Users trying to interact with OpenClaw as a spoken assistant

Severity:

• Annoying / UX friction, not data loss

Frequency:

• Every assistant reply with TTS enabled

Consequence:

• Extra manual click per response
• Voice mode feels like an attachment preview instead of live assistant speech

Evidence/examples

Observed in the OpenClaw Control app:

• TTS generated successfully using Microsoft provider / Edge TTS
• MP3 attachment appeared in the UI
• The embedded player initially showed 0 length/no volume controls while loading
• Audio played correctly once manually started
• Assistant voice was not heard automatically unless the user pressed play

Related config:

{
  "messages": {
    "tts": {
      "auto": "always",
      "provider": "microsoft"
    }
  }
}

### Additional information

This should remain opt-in to respect browser/app autoplay restrictions and user expectations around unexpected sound.