Midi Music Generator

Midi Music Generator

Composes MIDI music with AI music models, in the browser or on your own computer. For musicians, hobbyists and developers who want machine-made tunes in a format they can edit.

Made by skytnt 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: Midi Music Generator is an AI music composer that works with MIDI, the note format used by music software. The Space names six models for it, including Music-Llama and a set of midi-model files from skytnt, one of which is a J-pop LoRA.

The repository holds Python files for a MIDI model, a tokenizer and a synthesizer, plus a javascript folder and an example folder.

Who it suits: Musicians and hobbyists who want tune ideas in MIDI, and developers who want to read how a MIDI model is built.

Industry | Who uses it | How they use it | Example
Music | Composers and producers | Get melody ideas to edit in their music software | Generate a tune and open the MIDI in an editor
Games | Indie developers | Make placeholder background music | Generate a few short pieces to test in a build
Research | Music AI researchers | Study MIDI language models | Compare the named models on the same prompt

You can skip it if you need finished audio mastering, since the output is MIDI notes.

Where to use it:

• A web browser, on the Space's page at Hugging Face, with nothing to install
• Windows, macOS or Linux with Python, from a copy you download
• Docker, since the repository includes a Dockerfile

Use case | Who | What they get back
Compose a tune | Musicians | A generated piece in MIDI form
Try the example files | Newcomers | Results from the bundled examples
Study the models | Developers | Code for a MIDI model, tokenizer and synthesizer

How to set it up:

• Python on your computer
• The Python packages listed in requirements.txt
• The system packages listed in packages.txt, which the Space installs with apt-get
• The huggingface_hub package, which gives you the hf command

1. Install the huggingface_hub package.
2. Download the Space with the hf command.
3. Move into the downloaded folder.
4. Install the pinned Gradio version.
5. Install the packages from requirements.txt.
6. Start the ONNX version of the app with Python.

pip install -U huggingface_hub
hf download skytnt/midi-composer --repo-type space --local-dir midi-composer
cd midi-composer
pip install gradio==5.9.1
pip install -r requirements.txt
python app_onnx.py

How to use it: Open the Space's page in your browser, or start your own copy with the command below. Then use the controls on the page to compose a piece and listen to or save the result.

python app_onnx.py

Example: You open the Space and ask it to compose a piece. It generates a MIDI sequence, and you can take that MIDI into your own music software to edit or play it.

For newcomers and experts: If you are new to MIDI, open the page and generate one short piece with the default settings. Listen to it, then change one setting at a time to hear the difference.

If you know the field, the Space has two entry files, app.py and app_onnx.py, and the ONNX one is the main file. Several MIDI language models are named, so you can compare model sizes and styles. The Space runs on GPU hardware at Hugging Face.

You open the Space or run app_onnx.py
The MIDI tokenizer turns notes into tokens
A named model generates new note tokens
The result comes back as MIDI

Good to know:

• License: the Space lists apache-2
• Cost to run: none named
• Data it reaches: the settings and any MIDI you give it, plus the six models it names and the files in the Space
• Not for: finished audio production, since the output is notes in MIDI form

On Hugging Face, 13 November 2026

• Likes: 604
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/skytnt/midi-composer
• Files: https://huggingface.co/spaces/skytnt/midi-composer/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.