ComfyUI custom nodes for ACE-Step 1.5 - the most powerful open-source music generation model. Generate AI music directly in ComfyUI with commercial-grade quality.

• ACE-Step Model Loader - Load ACE-Step DiT and LLM models
• Generation Parameters - Configure music generation settings (BPM, duration, key, time signature)
• Music Creator - Generate music from text descriptions and lyrics
• Artist Node - AI-powered lyrics and music composition assistant with 50+ language support

1. Clone this repository into your ComfyUI custom_nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/HM-RunningHub/ComfyUI_RH_ACE-Step.git

2. Install dependencies:

cd ComfyUI_RH_ACE-Step
pip install -r requirements.txt

3. Restart ComfyUI

All models must be placed in ComfyUI/models/ACE-Step/ with the following structure:

ComfyUI/
└── models/
    └── ACE-Step/
        ├── vae/                          # VAE model (required)
        ├── Qwen3-Embedding-0.6B/         # Text embedding model (required)
        ├── acestep-v15-turbo/            # DiT model (required)
        └── acestep-5Hz-lm-1.7B/          # LLM model (choose one)
        └── acestep-5Hz-lm-4B/            # LLM model (optional, better quality)

# Download main model package (includes all required models)
huggingface-cli download ACE-Step/Ace-Step1.5 --local-dir ComfyUI/models/ACE-Step

# Optional: Download 4B LLM model for better quality
huggingface-cli download ACE-Step/acestep-5Hz-lm-4B --local-dir ComfyUI/models/ACE-Step/acestep-5Hz-lm-4B

# Install modelscope CLI
pip install modelscope

# Download main model
modelscope download --model ACE-Step/Ace-Step1.5 --local_dir ComfyUI/models/ACE-Step

Download the example workflow from workflows/example_workflow.json and import it into ComfyUI.

The example demonstrates two generation modes:

1. Artist Mode - Simple prompt: "一首中国风伤感恋爱歌曲" (A Chinese-style sad love song)
2. Manual Mode - Full control with custom caption, lyrics, BPM, duration, key, and time signature

1. Loader Node - Load ACE-Step models (DiT + LLM)
   • Select LLM type: acestep-5Hz-lm-1.7B or acestep-5Hz-lm-4B
2. GenerationParams Node - Set music parameters:
   • Caption: Music style description
   • Lyrics: Song lyrics (with structure tags like [Verse], [Chorus])
   • BPM: Beats per minute (30-300)
   • Duration: Length in seconds (10-600)
   • Key Scale: Musical key (e.g., "B minor")
   • Time Signature: Beats per measure (e.g., "4")
3. Creator Node - Generate the music
4. SaveAudio Node - Save the generated audio

Use the Artist Node for AI-assisted composition:

• Input a simple prompt describing the song you want
• LLM automatically generates caption, lyrics, and music parameters
• Supports 50+ languages for vocals
• Optional instrumental mode (no vocals)

Auto Detect, Arabic, Azerbaijani, Bengali, Bulgarian, Catalan, Chinese, Czech, Danish, Dutch, English, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Spanish, Swedish, Thai, Turkish, Ukrainian, Vietnamese, Cantonese, and more (50+ languages).

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

• ACE-Step 1.5 Official - Original ACE-Step project
• HuggingFace Models - Model downloads
• ComfyUI
• RunningHub

This project is based on ACE-Step 1.5, co-led by ACE Studio and StepFun.