ACE-Step 1.5

ACE-Step 1.5

Generates full songs from a text description, with a language model that plans lyrics and structure and a diffusion transformer that renders the audio. Built to run on consumer hardware.

Made by ACE-Step on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: ACE-Step 1.5 is an open music generation model from ACE-Step, built to run on consumer hardware. You describe the song you want and it produces the audio. A language model first plans the song, writing metadata, lyrics and captions, and a diffusion transformer then renders the music. The work comes from ACE Studio and StepFun together.

Who it suits: Musicians, producers, content creators and developers who want to make music from text on their own computer.

Industry | Who uses it | How they use it | Example
Music production | Producers | Sketch song ideas and loops from a text description | A short loop for a beat
Content creation | Video makers | Make background music for clips | A backing track for a video
Software | Developers | Add music generation to an app | A button that generates a track
Research | ML researchers | Study or fine-tune an open music model | A comparison of the base and turbo models

If you only want to listen to music rather than make it, you can skip it.

Where to use it:

• Windows, macOS or Linux with Python, for downloading the files
• Python programs that use the transformers library
• A browser, through the demo Space the card links

Use case | Who | What they get back
Text to music | Creators | A full song from a short description
Cover generation | Producers | An existing track redone in a new style
Repainting | Producers | A chosen section of a track regenerated
Vocal to BGM | Producers | Background music built from a vocal track

How to set it up:

• Python, to run the Hugging Face download tool
• A consumer graphics card, which the card says needs under 4GB of video memory
• Space for the model files

1. Install the Hugging Face hub tools with pip.
2. Download the model files with the hf command.
3. Follow the model card to run the model.

pip install -U huggingface_hub
hf download ACE-Step/Ace-Step1.5

How to use it: You give the model a plain description of the song. The language model turns it into a detailed plan, and the diffusion transformer makes the audio. The card says plans range from short loops to 10-minute pieces.

Tool | What it does
acestep-v15-base | Runs 50 steps with guidance and supports every task in the card's table
acestep-v15-sft | Runs 50 steps and is rated high on quality
acestep-v15-turbo | Runs 8 steps without guidance and is rated above sft on quality
acestep-5Hz-lm | The planning language model, in 0.6B, 1.7B and 4B sizes

Example: You type "an upbeat acoustic pop song about a road trip, with a sing-along chorus". The language model drafts lyrics, a title and style notes, and the diffusion transformer returns the finished track.

For newcomers and experts: Start with the turbo model, which needs only 8 steps, and ask for a short song in one style. The card's table links this repository from its turbo row.

The language model acts as a planner that writes a song blueprint through chain-of-thought, and that blueprint guides the diffusion transformer. The card says the alignment uses reinforcement learning based only on the model's own mechanisms, not external reward models or human preferences. The 4B planner is rated strong on composition.

Write a short song description
Language model plans lyrics and style
Diffusion transformer renders audio
A finished track comes back

Good to know:

• License: MIT
• Cost to run: none named, though you need a computer with a graphics card
• Data it reaches: downloads files from Hugging Face, then runs where you load it
• Not for: the turbo-rl model, which the card lists as not yet released

Title: Inference steps per DiT model
acestep-v15-base: 50 steps
acestep-v15-sft: 50 steps
acestep-v15-turbo: 8 steps

On Hugging Face, 13 November 2026

• Task: text to audio
• Library: transformers
• Formats: Safetensors
• Downloads in the last 30 days: 98,086
• Likes: 952
• License: mit
• Quantized versions: 18
• Fine-tunes: 27
• Tags: transformers, diffusers, safetensors, acestep, feature-extraction, audio, music, text2music, text-to-audio

Links

• Model: https://huggingface.co/ACE-Step/Ace-Step1.5
• Files: https://huggingface.co/ACE-Step/Ace-Step1.5/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download ACE-Step/Ace-Step1.5

Then follow the usage in its model card. Nothing in this kit runs by itself.