Kiku ACE-Step 1.5

Kiku ACE-Step 1.5

Make music with the ACE-Step 1.5 model: songs from text and lyrics, instrumentals, remixes, repaints and a band written around your own vocal. It is the engine behind the Kiku AI studio.

Made by Reubencf on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space runs the ACE-Step 1.5 music model and exposes it as a set of endpoints. The Kiku AI studio sends requests to it. The inputs and outputs of its endpoints match those of the Kiku-Engine Space. That lets the studio route a request to either one.

The model wants Python 3.11 along with Gradio 6 and torch 2.9. Kiku-Engine's older stack cannot hold them, so this Space stands alone.

Who it suits: It suits music makers who want AI help with songs, covers and band parts, and developers who are building a studio that talks to music models.

Industry | Who uses it | How they use it | Example
Music production | Songwriters | Send a text description and lyrics to get a song | Sketching a demo from a lyric sheet
Video and games | Content creators | Ask for an instrumental to fit a project | Getting a background track from a text description
Music production | Producers | Restyle, repaint or extend an existing recording | Changing the style of a take while keeping its tune
AI development | Developers | Call the endpoints from their own studio | Switching between this Space and Kiku-Engine

Skip it if you want a simple one-box music generator, since this Space is an engine for the Kiku AI studio.

Where to use it: The Space runs in these places.

• The Kiku AI studio, which sends requests to this Space
• A web browser, on the Space's page at Hugging Face
• A local copy on Windows, macOS or Linux, running Python 3.11
• Hugging Face's zero-a10g hardware, where the online Space runs

Use case | Who | What they get back
Write a song from lyrics | Songwriters | A full song planned by the language model
Restyle or extend a track | Producers | A new version of the audio
Add a band to a vocal | Singers | Drums, bass, keyboard and guitar mixed under the vocal

How to set it up: Online you only need a browser. A local copy needs these.

• Python 3.11 on Windows, macOS or Linux
• The pip command
• The system packages listed in packages.txt, which Hugging Face installs with apt-get
• Room on disk for the ACE-Step models

1. Install the huggingface_hub package.
2. Download the Space's files into a folder with the hf command.
3. Move into that folder.
4. Install gradio at version 6.2.0.
5. Install the packages listed in requirements.txt.
6. Start the app with python app.py.

pip install -U huggingface_hub
hf download Reubencf/Kiku-ACE-Step-15 --repo-type space --local-dir Kiku-ACE-Step-15
cd Kiku-ACE-Step-15
pip install gradio==6.2.0
pip install -r requirements.txt
python app.py

How to use it: The Kiku AI studio sends requests to the endpoints below. You can also open the Space's page and use them there. The remix endpoint has a strength setting for how closely it keeps the source.

Tool | What it does
/song | Makes a song from a text description and lyrics, planned by a 1.7B language model
/instrumental | Makes music from a text description alone
/remix | Restyles an existing track, with a strength setting for how close to stay
/repaint | Redoes a part inside the audio
/extend | Continues the audio past its end
/part | Writes one instrument into a mix, using the base model
/accompany | Writes a band around a vocal, using the base model

Example: You send /song a text description of a mood and a set of lyrics. The Space plans the song with its language model and returns the music.

For newcomers and experts: If you are new, open the Space's page and try /instrumental with a short description, since it needs no lyrics. Then try /song with a few lines.

If you know the field, note that /accompany throws away any vocal the model generated itself. First Demucs splits the singer from any backing already on your take. Next the base model composes drums, bass, keyboard and guitar to fit that vocal, listening to the audio rather than a tempo in the prompt. Last, Demucs strips any singing the model produced, and the band goes under your own take.

Send a request to an endpoint
Demucs separates the voice if needed
ACE-Step 1.5 writes or reworks the music
The result returns to the studio

Good to know: The acestep folder holds code taken from ACE-Step's own Space, released under MIT. The license text sits in a file named LICENSE.acestep. The copy drops the interface, training and nano-vllm parts. The README says the results may be used commercially.

• License: MIT
• Cost to run: none named
• Data it reaches: the audio and text you send, plus the ACE-Step model repositories
• Not for: training models, since the copied code leaves out the training parts

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: mit

Links

• Space: https://huggingface.co/spaces/Reubencf/Kiku-ACE-Step-15
• Files: https://huggingface.co/spaces/Reubencf/Kiku-ACE-Step-15/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.