ace-step-inference

Coordinate ACE-Step inference to generate stereo 48kHz audio from prompts or reference tracks.

1|Updated Nov 24, 2025
One-click install
npx skills add https://github.com/SpiralCloudOmega/DevTeam6 --skill ace-step-inference
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ace-step-inference
Source: https://github.com/SpiralCloudOmega/DevTeam6/tree/main/.github/skills/ai-integration/ace-step-inference
Command: npx skills add https://github.com/SpiralCloudOmega/DevTeam6 --skill ace-step-inference

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

ACE-Step-1.5 music generation integration enables producing full stereo audio from text prompts, reference tracks, or partial audio inputs, streamlining creative workflows in DAWs.

Core Features & Use Cases

  • Text-to-Music generation from prompts
  • Style transfer with reference audio (covers)
  • Repainting/inpainting on existing audio
  • CUDA-accelerated GGML backend with 48kHz output
  • JUCE integration for asynchronous generation

Quick Start

Start a 30-second ACE-Step inference from a text prompt and optional reference audio to generate a stereo 48 kHz track.

Frequently Asked Questions about ace-step-inference

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate 48kHz stereo music from text prompts in my DAW?

Text-to-music generation from prompts produces full stereo 48kHz audio by coordinating ACE-Step inference with a CUDA-accelerated GGML backend directly within DAW workflows.

Can I use a reference track to create style transfer covers with CUDA acceleration?

Yes, style transfer with reference audio creates covers by applying CUDA-accelerated ACE-Step inference to generate stereo audio matching the reference track characteristics.

Does ACE-Step inference support repainting and inpainting on existing audio?

ACE-Step inference supports repainting and inpainting on existing audio, allowing partial audio inputs to be modified while generating full stereo 48kHz output through the GGML backend.

Do I need CUDA to run the GGML backend for music generation?

CUDA acceleration is required for the GGML backend used in ACE-Step music inference, enabling model loading and asynchronous generation via a JUCE-based thread and Python API bridge.

What are the limitations of asynchronous music generation in JUCE?

Asynchronous generation runs via a JUCE-based thread to prevent audio blocking, but output is limited to 48kHz stereo and depends on CUDA-accelerated model loading with quantized GGML backends.