Environment

• macOS Companion app: v2026.2.12 (build 9500), MacBook Pro
• Audio output: Tested with both Bluetooth AirPods and built-in speakers
• Gateway: OpenClaw 2026.2.12 on Mac Studio (ms4), remote mode via wss://

Talk Config

• voiceId=bXHmd8SGNtE1r0uU8taf
• modelId=eleven_v3
• apiKey configured
• interruptOnSpeech=true

Symptoms

ElevenLabs TTS API call succeeds (correct voice ID, audio returned in ~300ms), but audio playback fails immediately. PCM playback fails first, MP3 retry also fails, then falls back to macOS system voice (AVSpeechSynthesizer). Happens with both pcm_44100 (default) and mp3_44100_128 output formats.

Log Evidence

talk TTS request voiceId=bXHmd8SGNtE1r0uU8taf chars=102
talk TTS synth timeout=20.0s
talk audio result finished=false interruptedAt=nil
talk TTS failed: audio playback failed; falling back to system voice
talk system voice start chars=102

Key Observation

The finished=false and interruptedAt=nil appear ~300ms after the TTS request, suggesting the audio player fails on the first frame rather than mid-stream.

What I Verified

• ElevenLabs API works: curl test from the gateway host returns valid MP3 audio
• Config is read correctly: Companion app logs show the correct voiceId, modelId, and apiKey
• Not a Bluetooth issue: Same failure occurs with built-in MacBook speakers
• Both output formats fail: pcm_44100 (default) and mp3_44100_128

Workaround

None currently. The system voice fallback works but uses the wrong voice.