sag

The sag skill provides a professional-grade text-to-speech (TTS) interface for OpenClaw, bridging ElevenLabs advanced neural synthesis with the intuitive command-line style of the macOS 'say' utility. Designed for users who demand high-fidelity audio output for their personal AI assistant, this skill allows for natural, human-like voice responses in various tones and delivery styles. It is ideal for developers and power users who want to integrate expressive voice capabilities into automated workflows, chat interactions, or personalized audio feedback systems.

• Full support for ElevenLabs models including v3 (expressive), multilingual v2, and fast v2.5 variants.

• Mac-style CLI command structure for rapid audio generation and local playback.

• Native support for voice prompting and character-specific delivery rules, including custom prosody like whispers, shouts, laughter, and sighs.

• Integration ready: generates audio files for integration with chat platforms via the OpenClaw media pipeline.

• Configurable audio delivery parameters including normalization for numbers, units, and URLs, as well as language bias settings to improve accuracy.

• Ensure ELEVENLABS_API_KEY or SAG_API_KEY is configured in your environment for authentication.

• Use the [tag] syntax (e.g., [whispers], [excited]) at the start of text lines to influence model delivery for creative characters like 'crazy scientists' or 'dramatic storytellers'.

• Utilize the --normalize auto flag for reliable handling of URLs and numerical data, or disable it when processing complex proper nouns.

• For chat responses, specify the voice using the -v flag or set a default voice ID (SAG_VOICE_ID) to maintain a consistent persona.

• Compatible with SSML-like features (depending on the model version) and supports pausing via [pause], [short pause], or [long pause] tags in v3 models.