test-voice-samplelisted

# Test Voice Sample / 试听样本 ## When to use / 何时使用 Immediately after cloning or tuning. Cheap sanity check before spending on full batch. 克隆或调参后立即使用，廉价验证，避免批量浪费。 ## Inputs / 输入 - `config/voice.json` (`voice_id`, current `stability`, `similarity_boost`). - Optional `--text` (default: pulls a representative line from the script). ## Output / 输出 - `workspace/voice/samples/test_<timestamp>.mp3`. - `workspace/voice/samples/test_<timestamp>.json` with params used. ## Steps / 步骤 1. Pick test text: `--text` if given, else first sentence containing a glossary term. 2. Call `mcp__elevenlabs__tts_convert` with current params. 3. Save MP3 + sidecar JSON. 4. Print local file path for the user to audition. ## Examples / 示例 ``` claude run test-voice-sample --text "Welcome to Agentic Headshot Video Studio." ``` ## Failure recovery / 失败回退 - If API error, retry once with exponential backoff. - If audio length < 0.5s, treat as failure and surface raw error.