MusicLM - Pytorch
lucidrains/musiclm-pytorchReviewed by AI
Checked by RepoPicks Radar on
What is MusicLM - Pytorch?
This is a community rebuild, in PyTorch, of MusicLM, the Google model that turns a text description into music. It is a library, not a program: you install it with pip and write Python code. It ships no trained model, so it makes no music until you gather a large collection of audio with matching text and train several networks on a graphics card, which takes serious time and hardware. It is for machine learning researchers and developers; a person who just wants a song from a sentence cannot use it as it is.
MusicLM - Pytorch is free and open source (MIT) and can replace Suno and Udio.
How easy is it to install MusicLM - Pytorch?
E · developers only · Windows
- Windows E
- Linux E
On Linux, the usual system for this kind of training, it is developers only: `pip install musiclm-pytorch`, then write Python, prepare a dataset of sound and text pairs and train three transformers plus the MuLaN model before anything can be generated. Windows works the same way if you have an NVIDIA card, since the example code calls .cuda(); there is no app, no interface and no ready model to download.
Coming soon: install in 1 click with the RepoPicks app.
MusicLM - Pytorch on GitHub
- Stars
- 3,294on Nov 10, 2026
- Downloads
- —all releases
- Growth
- —no gain in the last reading
Stars over time
Facts
- Runs on
- Linux, Windows
- You need to install
- Python, Graphics card
- To build from source
- Python, Graphics card
- Disk space
- 950 MB–4.3 GB · estimate
- Accounts or keys
- None
- Cost
- Free · The code is free (MIT). The real cost is the hardware or rented GPU time needed to train the models.
- License
- MIT (Permissive)
- Last change on GitHub
- Sep 6, 2023
MusicLM - Pytorch risk signals (4)
- Attentionstale releases
Last release on 2023-09-06 (1158 days ago)
automatic check
- Infono trained model included
Nothing is ready to use: the README says "MuLaN first needs to be trained", "get a ton of <sound, text> pairs and train" and "After much training on all three transformers (semantic, coarse, fine), you will pass your finetuned or trained-from-scratch AudioLM". There are no model files in the repository and the releases have "no platform binaries".
AI review
- Inforeplication still incomplete
The project is a rebuild of someone else's model and its own Todo list still has open items: "support variable lengthed audio with masking in audio transformer", "add a version of mulan to open clip" and "set all the proper spectrogram hyperparameters". The README invites help: "join ... if you are interested in helping out with the replication with the LAION community".
AI review
- Infono longer updated
The last commit and the last release are both from 2023-09-06, and setup.py still marks it "Development Status :: 4 - Beta" at version 0.2.8.
AI review
Evidence is shown in its original language.
What Radar could not check about MusicLM - Pytorch
- Whether training it really produces music close to Google's published MusicLM examples
- How much audio data, GPU time and money a usable model would take
- Whether the pinned dependencies (audiolm-pytorch, x-clip, torch) still install together today
- Real disk use once a dataset and the training checkpoints are on disk
- Which rights apply to the music collection you would need for training