Voice: cloud Qwen3-TTS is the default engine in prod (was Kokoro)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

web/tts.js

@@ -10,8 +10,9 @@ import { playSamples, stopAudio } from '/web/ttsAudio.js'
 import { ensurePersistentStorage } from '/web/storage.js'
 
 const ENGINES = [kokoro, qwen3local, qwen3, kitten, webspeech]
-// On localhost the local-GPU Qwen3-TTS is the default; in prod it's Kokoro (local-first).
-let activeId = isLocalhost() ? 'qwen3local' : 'kokoro'
+// Default voice engine: local-GPU Qwen3-TTS on localhost, cloud Qwen3-TTS (DashScope)
+// in prod. (Kokoro/Kitten/Web Speech stay as the in-browser, local-first alternatives.)
+let activeId = isLocalhost() ? 'qwen3local' : 'qwen3'
 
 // Qwen3-TTS designs a voice from a free-form description (the persona's `voice`).
 // Panels set it before narrating; previewVoice() plays a one-off sample.