Wiki source code of Audiocraft AI

Last modified by René Schmidt on 2025/03/28 12:22
Show last authors
1 = AudioCraft =
2
3
4 [[~[~[image:https://github.com/facebookresearch/audiocraft/workflows/audiocraft_docs/badge.svg~|~|alt="docs badge"~]~]>>url:https://github.com/facebookresearch/audiocraft/workflows/audiocraft_docs/badge.svg]] [[~[~[image:https://github.com/facebookresearch/audiocraft/workflows/audiocraft_linter/badge.svg~|~|alt="linter badge"~]~]>>url:https://github.com/facebookresearch/audiocraft/workflows/audiocraft_linter/badge.svg]] [[~[~[image:https://github.com/facebookresearch/audiocraft/workflows/audiocraft_tests/badge.svg~|~|alt="tests badge"~]~]>>url:https://github.com/facebookresearch/audiocraft/workflows/audiocraft_tests/badge.svg]]
5
6 AudioCraft is a PyTorch library for deep learning research on audio generation. AudioCraft contains inference and training code for two state-of-the-art AI generative models producing high-quality audio: AudioGen and MusicGen.
7
8 == Installation ==
9
10
11 AudioCraft requires Python 3.9, PyTorch 2.1.0. To install AudioCraft, you can run the following:
12
13 {{{# Best to make sure you have torch installed first, in particular before installing xformers.
14 # Don't run this if you already have PyTorch installed.
15 python -m pip install 'torch==2.1.0'
16 # You might need the following before trying to install the packages
17 python -m pip install setuptools wheel
18 # Then proceed to one of the following
19 python -m pip install -U audiocraft # stable release
20 python -m pip install -U git+https://git@github.com/facebookresearch/audiocraft#egg=audiocraft # bleeding edge
21 python -m pip install -e . # or if you cloned the repo locally (mandatory if you want to train).
22 python -m pip install -e '.[wm]' # if you want to train a watermarking model}}}
23
24 We also recommend having ffmpeg installed, either through your system or Anaconda:
25
26 {{{sudo apt-get install ffmpeg
27 # Or if you are using Anaconda or Miniconda
28 conda install "ffmpeg<5" -c conda-forge}}}
29
30 == Models ==
31
32
33 At the moment, AudioCraft contains the training code and inference code for: