Audiocraft Tutorial: Create Music Easily with Meta's AI Tool

What is Audiocraft?

On June 9, 2023, Meta announced the launch of Audiocraft, an innovative AI tool that specializes in music generation and audio processing, powered by advanced deep learning technology. Unlike Google's MusicLM, Audiocraft stands out as an open-source platform, giving users the flexibility to explore and create without constraints.

Prerequisites for Installation

Python 3.9 or later
pip package manager
PyTorch 2.0
ffmpeg

Before installing PyTorch, you can use the following command in your terminal:

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113

To install ffmpeg, run the command:

brew install ffmpeg

Installation of Audiocraft

To get started with Audiocraft, clone the GitHub repository and navigate to the downloaded folder:

git clone https://github.com/facebookresearch/audiocraft.git

After that, install the required libraries by running:

pip install -r requirements.txt

With the installation complete, you are ready to start creating music!

Creating Music with Audiocraft

To utilize MusicGen, the specific module for music creation, Meta offers a user-friendly Gradio UI. Launch it by entering the following command:

python app.py

This will start the UI at http://127.0.0.1:7860/, where you can interact with the application and even generate a public URL for collaboration.

Choosing Your Model

Audiocraft allows you to select from four different models, which significantly influences the music output:

facebook/musicgen-melody: Generates music with melodies based on an uploaded music file.
Others: Produce more ambient music using text inputs only.

As you begin experimenting, expect the initial setup to take more time, especially if you're lacking a powerful GPU.

Understanding Duration Limits

The default maximum song duration is capped at 120 seconds. However, you can modify this value to create longer pieces. Open the musicgen_app.py file in the demos folder and locate the duration variable (typically around line 240). Modify it to your desired length, save, and restart the program to implement the change.

Generating a full-length song (e.g., 3:20 minutes) on a machine with 16GB RAM may take significant time, so patience is key.

Integrating Audiocraft into Your Project

Now, let’s explore how to integrate Audiocraft into an existing codebase. Here’s a simple approach to create an app that generates music based on a song description derived from a URL.

Step-by-Step Guide

Create a new folder named audiocraft_app.
Within this folder, create a file called audiocraft_app.py and a requirements.txt file.
Fill the requirements.txt file with necessary libraries.
Install the required libraries by running:

pip install -r requirements.txt

Write the necessary code in the audiocraft_app.py file.

Run the app with:

streamlit run audiocraft_app.py

Your app should be accessible at http://localhost:8501/. Insert your OpenAI API Key, select the appropriate ChatGPT model based on the content length, and provide the URL.

Conclusion

Audiocraft represents an exciting frontier in AI-powered music generation. Its flexibility and open-source nature offer countless possibilities for musicians and developers. As you dive into creating with Audiocraft, don’t forget to share your musical masterpieces!