Agent skill · hairyf
sonic
AI voice (TTS), music, lyrics, and sound-effect generation for agents. Use when the user asks to generate, synthesize, or create speech, voiceovers, songs, music, background music, lyrics, or sound effects. Provides a zero-dependency Node.js script covering SiliconFlow / MiniMax / OpenAI / ElevenLabs TTS, MiniMax music, and local Sony Woosh / MMAudio sound effects.
What it needs
About 2k tokens when loaded.
What this skill does
Sonic Generate voice, music, lyrics, and sound effects through scripts/sonic.js — a zero-dependency Node.js CLI (Node 18+, no npm install). It saves audio to disk and prints the absolute path. Quick start Model selection China-friendly priority order (Alipay/WeChat top-up, no foreign card): Asset Recommended (priority order) Notes ------- ------------------------------ ------- Voice / TTS SiliconFlow CosyVoice2 → MiniMax Speech → OpenAI TTS / ElevenLabs SiliconFlow is OpenAI-compatible and already keyed Music MiniMax Music 3.0 → Bailian Fun-Music → Suno (third-party relays) MiniMax has an official API + China payment Sound effects Local MMAudio → Local Sony Woosh → MMAudio cloud → ElevenLabs SFX Local channels run on an NVIDIA GPU, free, auto-installed by the agent Key workflows Local SFX prompts must be English — MMAudio and Woosh use English CLIP/CLAP text encoders; non-English prompts produce garbled, voice-like artifacts. Add --negative "speech, voice, human talking, vocals, singing" to suppress them. Music prompts: {genre/style} + {instruments} + {mood} + {tempo} + {scene}; --instrumental for pure music, --lyrics-optimizer or --lyrics for vocals. Generation takes tens of seconds to minutes — use a generous timeout. TTS: --speed 0.9–1.1 for narration, 1.2+ for ads, 0.8 for audiobook; -v <voice> keeps a consistent voice. ElevenLabs' free tier only ships its 21 premade voices (default "Adam") — list them with GET /v1/voices. Local channels auto-install: when local Woosh / MMAudio is needed but missing, run scripts/install/woosh. / scripts/install/mmaudio. with -Start / --start yourself — never ask the user to install by hand. Servers idle-exit after 15 minutes; restart before retrying a failed request. Woosh's API server serves Woosh-DFlow (fast) only; quality and video-to-audio modes need the Gradio UI. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills hairyf/sonic