README.md

FluxMusic: Text-to-Music Generation with Rectified Flow Transformer
_{GUI Implementation}

   

This repo contains a Graphical User Interface (GUI) implementation of the FluxMusic model, based on the paper Flux that plays music. It explores a simple extension of diffusion-based rectified flow Transformers for text-to-music generation.

FluxMusic GUI

<img src=visuals/FluxMusicGUIsc.png width=600 />

I created a user-friendly GUI for FluxMusic using Gradio. This interface allows users to easily generate music based on text prompts without needing to interact with command-line interfaces.

Features:

1. Model Selection: Users can choose from different FluxMusic models (small, base, large, giant) via a dropdown menu.

2. Text Prompt: Enter your desired text prompt to guide the music generation.

3. Sliders and Inputs:

   • Seed: Set a seed for reproducibility (0 for random).
   • CFG Scale: Adjust the Classifier-Free Guidance scale (1-40).
   • Steps: Set the number of diffusion steps (10-200).
   • Duration: Specify the length of the generated audio in seconds (10-300).

4. File Management:

   • Models Folder: Place your FluxMusic model files (.pt) in the models folder.
   • Generations Folder: Generated audio files are saved in the generations folder.

5. File Naming System: Generated files are named using the format: [prompt]_[seed]_[model]_[counter].wav

Setup and Running

1. Install the required dependencies:

   pip install -r requirements.txt
   

2. Place your FluxMusic model files in the models folder.

3. Run the GUI:

   python fluxGUI.py
   

4. Use the interface to generate music based on your prompts and preferences.

Model Information

FluxMusic comes in four sizes: Small, Base, Large, and Giant. You can download these models from the following links:

Model	Url
FluxMusic-Small	link
FluxMusic-Base	link
FluxMusic-Large	link
FluxMusic-Giant	link

Acknowledgments

The codebase is based on the awesome Flux and AudioLDM2 repos.