jarvis-voice
由 globalcaos 提供
Speech & Transcription简介
Metallic AI voice persona with TTS and visual transcript styling.
描述
该技能的功能
Adds an offline British-accented voice to an OpenClaw agent using sherpa-onnx TTS (Alan voice) with a metallic effects chain via ffmpeg. Audio plays in the background before the text response appears on screen. Four humor patterns shape the agent's conversational personality: dry wit, self-aware AI, alien observer, and literal idiom play.
Runs fully offline with a custom audio effects chain, so voice output requires no cloud APIs, network access, or per-request costs.
何时使用
- Hearing a spoken greeting at the start of each work session
- Getting voice readouts of task results without looking at the screen
- Running a hands-free terminal assistant while doing other work
- Receiving spoken summaries after long background operations finish
- Using a consistent AI persona with a recognizable voice and tone
示例工作流
User asks: 'What did you finish while I was away?'
- 1
Runs
jarvis "Here is a summary of completed tasks, sir."with background: true before writing any text - 2
User hears the British Alan voice with pitch shift, flanger, and echo effects before text appears
- 3
Writes the visible transcript using the bold **Jarvis:** prefix in the chat response
- 4
Webchat post-render hook auto-wraps the transcript text in a purple italic span
User hears the spoken summary first, then sees the styled transcript on screen
前置要求
- SHERPA_ONNX_TTS_DIR environment variable pointing to the sherpa-onnx TTS installation directory
安装
clawhub install globalcaos/jarvis-voice运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。