jarvis-voice_skill

Overview

This skill gives your OpenClaw agent a JARVIS-inspired metallic voice with fully offline TTS using sherpa-onnx and local audio processing. It produces purple, italicized transcripts in webchat and applies ffmpeg effects (flanger, echo, pitch shift) for a robotic sheen. The pipeline is local-first, with zero cloud calls, zero API costs, and minimal latency. Playback and speed are configurable for efficient interactions.

How this skill works

Text is synthesized locally with sherpa-onnx (VITS/Piper recommended), then processed through ffmpeg to add metallic effects like echo, chorus, and high/low filters. The script plays audio through ALSA (aplay) and can run at 2x speed for faster delivery. Webchat displays a styled transcript span so spoken lines are visually distinct (purple, italic). Parameters live in the jarvis script so voice presets and playback device are configurable.

When to use it

• You want an offline TTS voice with a distinct JARVIS-style persona.
• You must avoid cloud APIs and keep audio processing local for privacy or cost reasons.
• You need a visual transcript in chat that matches spoken audio (styling supported).
• You want tunable metallic effects (echo, chorus, pitch) without external services.
• You need fast playback for quicker agent interactions.

Best practices

• Install sherpa-onnx with a recommended VITS model (en_GB-alan-medium suggested) for best timbre.
• Configure your ALSA device in the jarvis script (aplay -D plughw:X,Y) before use.
• Keep voice effect parameters in version control and document presets for reproducibility.
• Start with milder effects and incrementally increase echo/chorus to avoid unintelligible audio.
• Allow span tags and add the provided CSS (.jarvis-voice) to webchat sanitization and styles.

Example use cases

• Personal assistant that speaks alerts and shows purple italic transcripts in chat.
• Local kiosk or offline agent in secure environments where no cloud TTS is allowed.
• Developer testing: fast, local TTS to evaluate conversational flows without API costs.
• Accessibility mode: offer spoken responses alongside readable, styled transcripts.
• Roleplay or demo where a metallic JARVIS persona enhances user experience.