Proof of Use. An independent field guide · No. 01
Back to the collection
Music & Audio

AudioCraft / MusicGen

Explore text- and melody-conditioned music generation with Meta’s AudioCraft research toolkit and MusicGen models.

music generationMusicGenmelodyresearchnoncommercial weights
Source-reviewedNot hands-on testedSources checked

Start with an idea

What you can make

01

Sketch instrumental musical ideas for noncommercial experimentation

02

Study how prompts and a permitted reference melody influence generated music

Creator demonstration

See the documented workflow

Official project sample page linked by the repository, including text-conditioned and stereo examples; curated examples are not an independent quality test.

Hear the creators’ MusicGen samples

Before you begin

What it takes to run

Python + GPU setup

  • Repository installation instructions specify Python 3.9 and PyTorch 2.1.0
  • FFmpeg is recommended; check the pinned dependency stack before choosing an environment
  • Download model checkpoints; allow GPU memory and runtime for the chosen model and clip duration
Runs through / on

Python / PyTorch · GPU workstation · Local Gradio interface · Official browser demo

The real tradeoffs

Cost & licensing

Running costs

No software fee for local use within license terms; GPU compute is your cost.

The official MusicGen guide calls for a GPU and at least 16 GB of memory for medium models. Smaller models or shorter clips may fit smaller GPUs; rented hardware and hosted demo availability are separate.

Code license

MIT

Read the code license

Models & other terms

CC-BY-NC-4.0

Read the model terms

Code and supplied model weights have different licenses. The noncommercial weights restriction means this is not an unrestricted commercial-generation option.

Know the limitations

  • MusicGen does not generate realistic vocals
  • English prompts are best supported; styles and cultures are represented unevenly
  • Outputs can collapse to silence, and prompt experimentation may be necessary
  • A research toolkit with noncommercial model weights needs additional review before downstream deployment

Follow the evidence

Sources & verification

Official sources were reviewed on October 14, 2026. We did not install, run or benchmark this project. This is a source-check date, not a release date.

Links and project terms can change. These sources document the claims above; they are not a safety certification. Read our full method.

Keep exploring

A few other possibilities

All tools

05Music & Audio

Whisper

Transcribe speech in multiple languages or translate spoken audio into English with OpenAI’s locally runnable speech-recognition models.

Try it for

Create a reviewable transcript of your own interviews or podcasts

Read entry Official examples
Setup · Moderate
Python + FFmpeg
Cost to consider
Local compute

07Music & Audio

Chatterbox

Generate expressive speech from text with Resemble AI’s local text-to-speech family, including multilingual and reference-voice options.

Try it for

Prototype narration or spoken interfaces from your own scripts

Read entry Official examples
Setup · Moderate
Python + model download
Cost to consider
Local or rented compute

08Music & Audio

Demucs

Separate a mixed music recording into vocals, drums, bass and other accompaniment with a locally runnable source-separation model.

Try it for

Isolate parts from your own recordings for practice and analysis

Read entry Official examples
Setup · Moderate
Python + audio dependencies
Cost to consider
Local compute; check weights terms