DeepJ
calclavia/DeepJReviewed by AI
Checked by RepoPicks Radar on
What is DeepJ?
DeepJ is the code behind a 2018 research paper on generating music in a specific mixture of composer styles. It is a set of Python scripts: one trains the model, another writes new MIDI files. There is no ready program, no trained model and no music data in the repository, so you have to collect your own MIDI collection and train it yourself on a machine with a graphics card. It is meant for people comfortable with Python and machine learning, and the code has not been touched since March 2018.
DeepJ is free and open source (MIT) and can replace Suno and Udio.
How easy is it to install DeepJ?
E · developers only · Windows
- Windows E
- Linux E
On Linux or Windows this is developers-only work: the README asks for Python 3.5, a separate library (python-midi) installed from source with setup.py, then pip install of Keras and tensorflow-gpu, and finally your own MIDI dataset placed in a data/ folder before any training can start. There is no installer, no release file and no trained model to simply run.
Coming soon: install in 1 click with the RepoPicks app.
DeepJ on GitHub
- Stars
- 744on Nov 10, 2026
- Downloads
- —all releases
- Growth
- —no gain in the last reading
Stars over time
Facts
- Runs on
- Windows, Linux
- You need to install
- Python, Graphics card
- To build from source
- Python, Graphics card
- Disk space
- 100 MB–300 MB · estimate
- Accounts or keys
- None
- Cost
- Free · MIT licensed code, free to use. You pay in your own hardware or rented GPU time for training.
- License
- MIT (Permissive)
- Last change on GitHub
- Mar 5, 2018
DeepJ risk signals (4)
- Attentionobsolete dependencies
README Requirements says "Python 3.5" (no longer supported since 2020) and requirements.txt lists unpinned "keras" and "tensorflow-gpu", a package name that TensorFlow stopped publishing, so a plain install on a current machine is unlikely to work without changes.
AI review
- Infono trained model or dataset
README: "The dataset is not provided in this repository. To train a custom model, you will need to include a MIDI dataset in the data/ folder." There are no release files and no weights in the top-level file list, so you cannot generate anything before training.
AI review
- Inforesearch prototype unmaintained
The project links to the paper arxiv.org/abs/1801.00887, calls the style control "a simple technique ... as a proof of concept", and the last commit is from 2018-03-05 with 17 open issues and no releases.
AI review
- Infothird party dependency from git
README tells you to clone github.com/vishnubob/python-midi and run "python3 setup.py install", so part of the install comes from another repository outside the Python package index.
AI review
Evidence is shown in its original language.
What Radar could not check about DeepJ
- Whether the code still installs and runs with today's Python, Keras and TensorFlow versions
- How long training takes, which graphics card it needs and how much memory and disk the dataset and checkpoints use (the 100-300 MB estimate is a guess, not measured)
- How good the generated music actually sounds, since the paper's rating comes from the authors
- Whether the external python-midi library still builds on a modern system