ACE-Step Skills for Any Agents

Custom Skills for AI Agents (Claude Code, OpenAI Codex, etc.) to generate music via ACE-Step API.

Available Skills

Skill	Description
acestep	Music generation via ACE-Step API
acestep-docs	Documentation and troubleshooting

Features

acestep (Music Generation)

• Text-to-Music - Generate music from descriptions
• Lyrics Generation - Auto-generate or manually specify lyrics
• Audio Continuation - Continue from existing audio
• Audio Repainting - Modify specific parts of audio
• Random Generation - Generate random music samples

acestep-docs (Documentation)

• Installation Guide - Setup and configuration help
• GPU Compatibility - VRAM requirements and hardware recommendations
• Gradio UI Guide - Web interface usage
• Inference Tuning - Parameter optimization
• API Reference - REST API and OpenRouter integration

Prerequisites

• ACE-Step API Server - Running ACE-Step V1.5 API service

Installation

Claude Code

Copy desired skill folders from skills/ to:

Project level (current project only):

your-project/.claude/skills/

Global level (all projects):

~/.claude/skills/

OpenAI Codex

Copy desired skill folders from skills/ to:

Project level:

your-project/.agents/skills/

Global level:

~/.agents/skills/

Directory Structure

skills/
├── acestep/                    # Music generation skill
│   ├── SKILL.md
│   └── scripts/
│       ├── acestep.sh
│       └── config.json
└── acestep-docs/               # Documentation skill
    ├── SKILL.md
    ├── getting-started/
    │   ├── README.md
    │   ├── Tutorial.md
    │   └── ABOUT.md
    ├── guides/
    │   ├── GRADIO_GUIDE.md
    │   ├── INFERENCE.md
    │   └── GPU_COMPATIBILITY.md
    └── api/
        ├── API.md
        └── Openrouter_API.md

Configuration (acestep)

Edit acestep/scripts/config.json to configure API connection and defaults:

{
  "api_url": "http://127.0.0.1:8001",
  "api_key": "",
  "generation": {
    "thinking": true,
    "use_format": true,
    "audio_format": "mp3",
    "vocal_language": "en"
  }
}

Options

Option	Default	Description
api_url	http://127.0.0.1:8001	API server address
api_key	""	API key (optional)
generation.thinking	true	Enable 5Hz LM (high quality)
generation.use_format	true	Enhance caption/lyrics via LM
generation.audio_format	mp3	Output format (mp3/wav/flac)
generation.vocal_language	en	Vocal language

Usage (acestep)

In Agent

After installation, mention music generation in conversation:

User: Generate a cheerful pop song
User: Create a song about spring
User: Generate jazz background music

Command Line

# Check API status
./scripts/acestep.sh health

# Generate music - Caption mode
./scripts/acestep.sh generate "Pop music with guitar"

# Generate music - Simple mode (LM auto-generates)
./scripts/acestep.sh generate -d "A cheerful song about spring"

# With lyrics
./scripts/acestep.sh generate -c "Lyrical pop" -l "[Verse] Hello world"

# Random generation
./scripts/acestep.sh random

# Check task status
./scripts/acestep.sh status <job_id>

Options

Option	Description
-c, --caption	Music style description
-d, --description	Simple description, LM auto-generates
-l, --lyrics	Lyrics content
--no-thinking	Disable thinking mode
--steps	Diffusion steps
--guidance	Guidance scale
--duration	Audio duration (seconds)
--bpm	Tempo

Output

Results saved to acestep_output folder:

project_root/
├── acestep_output/
│   ├── <job_id>.json      # Task result (JSON)
│   ├── <job_id>_1.mp3     # Audio file
│   └── ...

GPU Memory

VRAM	LM Model	Notes
≤6GB	None (DiT only)	LM disabled
6-12GB	acestep-5Hz-lm-0.6B	Lightweight
12-16GB	acestep-5Hz-lm-1.7B	Better quality
≥16GB	acestep-5Hz-lm-4B	Best quality

References

• acestep/SKILL.md - Music generation API documentation
• acestep-docs/SKILL.md - Documentation skill index
• ACE-Step - ACE-Step project