LIVE
Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|OpenAI models are coming to Google Cloud21/10/26 · OpenAI|OpenAI details ChatGPT's new advertising infrastructure21/10/26 · OpenAI|Liquid AI unveils LFM3, a base for specialized multimodal models27/10/26 · Liquid AI|US Police Departments Use AI to Flag 'Suspicious' Driving Behavior27/10/26|AllenAI questions whether LLMs can judge research novelty27/10/26 · AllenAI|Open-Source AI Tools Aim to Speed Up Dementia Research27/10/26 · NVIDIA|Solid-state transformers eyed as a fix for AI's power grid bottleneck27/10/26|OpenAI expands Oracle partnership to bring business apps into ChatGPT27/10/26 · OpenAI|OpenAI publishes a page announcing 'GPT-6.1 Terra', with no technical details available26/10/26 · OpenAI|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|OpenAI models are coming to Google Cloud21/10/26 · OpenAI|OpenAI details ChatGPT's new advertising infrastructure21/10/26 · OpenAI|Liquid AI unveils LFM3, a base for specialized multimodal models27/10/26 · Liquid AI|US Police Departments Use AI to Flag 'Suspicious' Driving Behavior27/10/26|AllenAI questions whether LLMs can judge research novelty27/10/26 · AllenAI|Open-Source AI Tools Aim to Speed Up Dementia Research27/10/26 · NVIDIA|Solid-state transformers eyed as a fix for AI's power grid bottleneck27/10/26|OpenAI expands Oracle partnership to bring business apps into ChatGPT27/10/26 · OpenAI|OpenAI publishes a page announcing 'GPT-6.1 Terra', with no technical details available26/10/26 · OpenAI|
AppWatch outofficial publisher

MusicGen (Meta)

App : A product you use directly

Meta's text-to-music generation model with open weights.

Landmark and easy to use, but weights are non-commercial.

Overall score
5.9 /10
Security: one caveat, see the card
Utility
7.0
Reliability
4.0
Security
One caveat
Ease
Technical · Technical

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

Composing music takes time, skills and expensive tools. MusicGen lets you generate musical excerpts directly from a text description, without playing an instrument or mastering production software.

MusicGen is part of Meta's AudioCraft library. It relies on a deep learning model trained on audio data: you enter a free-text description, and the model produces a corresponding audio file. A variant called MusicGen Style also accepts an audio clip as a style reference.

This tool is aimed at developers, researchers and technically minded creators who can set up a Python environment and work from the command line. Training code is also available for those who want to fine-tune the model on their own data.

MusicGen is not suited for users without technical experience: installation requires Python 3.9, PyTorch and ffmpeg. In addition, the model weights are released under the CC-BY-NC 4.0 licence, which prohibits any commercial use.

What it offers

  • ·Music generation from a free-text description.
  • ·MusicGen Style variant: conditioning on an audio clip provided as a style reference.
  • ·Training code available to fine-tune the model on your own data.
  • ·Integration within the AudioCraft library alongside other audio models (AudioGen, EnCodec, MAGNeT, JASCO).
  • ·musicgen-melody model available for melodic conditioning.
  • ·Built-in audio watermarking via AudioSeal to mark generated files.
  • ·Documented Python API to integrate MusicGen into your own projects.

In practice

Price
Free
Commercial use
Commercial use: no
Available on
linux, windows
Licence
MIT (code) / CC-BY-NC (poids)
Runs locally
Yes, 8 GB VRAM recommended
FR interface
No

Security and data

Rating
Security: one caveat, see the card
Your data
Data stays local
Maintenance
Inactive for over six months

Links and repository checked on 25 October 2026 · added on 1 June 2023

How to install MusicGen (Meta)

1. Make sure Python 3.9 is installed on your machine. 2. Install PyTorch 2.1.0 first: python -m pip install 'torch==2.1.0' 3. Install build tools: python -m pip install setuptools wheel 4. Install AudioCraft: python -m pip install -U audiocraft 5. Install ffmpeg: sudo apt-get install ffmpeg (Linux) or conda install "ffmpeg<5" -c conda-forge (Anaconda). 6. Model weights are downloaded automatically from Hugging Face on first use; make sure you accept the CC-BY-NC 4.0 licence before any commercial use.
python -m pip install -U audiocraft

In the same category

Whole category Audio, voice & music →