ACE-Step User-friendly Gradio Interface

This is a simplified and intuitive Gradio interface for ACE-Step, making AI music generation more accessible through a user-friendly UI.

Features

The interface offers three main capabilities:

1. Text to Music: Generate music from text prompts and lyrics
2. Audio Transformation: Transform existing audio files with new tags/lyrics or modify their style using denoising controls. Supports uploading files.
3. Song Editor: Edit sections of songs, add intros or outros

Getting Started

Installation

1. Make sure you have ACE-Step installed (https://github.com/ace-step/ACE-Step/)
2. Install additional dependencies:

   pip install gradio==4.26.0 soundfile
   

Running the App

Copy utils.py, gradio_app.py and param_maps.py into the original ACE-Step folder.

python gradio_app.py

Options

The app supports various command-line options:

Options:
  --checkpoint_path TEXT  Path to the checkpoint directory (default: auto-download)
  --bf16 BOOLEAN          Whether to use bfloat16 (default: True)
  --torch_compile BOOLEAN Whether to use torch compile (default: False)
  --device_id INTEGER     Device ID to use (default: 0)
  --share BOOLEAN         Whether to create a public share link (default: False)
  --server_name TEXT      Server name (default: 127.0.0.1)
  --server_port INTEGER   Server port (default: 7860)
  --help                  Show this message and exit.

Interface Guide

Text to Music

This tab focuses on straightforward generation from text prompts and lyrics.

1. Optionally select a Genre Preset to quickly populate the tags.
2. Enter Music Tags & Description: Use comma-separated tags describing the music you want (genre, instruments, mood, tempo, etc.)
3. Enter Lyrics: Structure your lyrics with tags like [verse], [chorus], and [bridge]. Use [instrumental] for instrumental music.
4. Adjust Song Duration: Set the length of the generated music in seconds.
5. Set Creativity / Variation: Lower values result in stricter adherence to the prompt; higher values allow more creative freedom. (This adjusts the guidance_scale).
6. Choose Generation Quality/Speed: Select between faster generation or higher quality output. (This adjusts infer_step and scheduler_type).
7. (Advanced) Manually set Guidance Scale values for fine-grained control.
8. (Advanced) Set Manual Seeds for reproducibility.
9. Click Generate Music to create your song.

Audio Transformation

This tab allows you to modify existing audio files using new descriptive tags and lyrics, controlling the amount of change.

1. Upload Input Audio: Upload an audio file to modify.
2. Enter New Tags: Describe the desired style/elements for the transformed audio.
3. Enter New Lyrics: Provide new lyrics or use [instrumental].
4. Adjust Denoise Amount: Control how much the original audio influences the output. Higher values lead to more significant transformation based on the new tags/lyrics.
5. (Advanced) Enable Advanced Settings to directly control n_min, n_max, infer_step, guidance_scale, cfg_type, and omega_scale for precise transformation control.
6. Click Transform Audio to create your modified version.

Note: The app automatically converts mono audio to stereo format if needed, as required by ACE-Step.

Song Editor

This tab lets you edit sections of songs, add intros, or outros.

1. Upload Song to Edit or click Load from Other Tab to edit a song generated in another tab.
2. Choose Edit Operation:
   • Inpaint Section: Replace a section of the song
   • Add Intro (Prefix): Generate music that leads into the song
   • Add Outro (Suffix): Generate music that extends the ending
3. Specify the timing parameters for your edit.
4. Enter Tags and Lyrics for the new section.
5. Adjust Variation Amount to control how different the new section is.
6. Click Apply Edit to Song to create your edited version.

Advanced Usage

Each tab contains an Advanced Options accordion offering more granular control:

• Text to Music: Manual guidance scales and seeds.
• Audio Transformation: Direct control over diffusion parameters (n_min, n_max, infer_step, etc.) and seeds.
• Song Editor: Manual seeds and retake variance.

Output

Generated audio appears in the right column of each tab. You can:

1. Play the audio directly in the browser
2. Download the audio file
3. View the generation parameters by expanding the Output Parameters accordion
4. Send the generated audio to the Song Editor tab for further editing

Project Structure

• gradio_app.py: Main application file
• param_maps.py: Module for mapping simplified UI controls to ACE-Step parameters
• utils.py: Utility functions for the app

How It Works

This Gradio interface simplifies ACE-Step's numerous parameters by:

1. Abstraction: Exposing intuitive controls (like Creativity, Denoise Amount) while managing complex underlying parameters.
2. Sensible Defaults: Using well-tested default values for many ACE-Step settings.
3. Parameter Mapping: Translating simple UI inputs (sliders, radio buttons) into appropriate combinations of ACE-Step parameters via param_maps.py.
4. Audio Processing: Automatically handling tasks like ensuring stereo audio input using functions in utils.py.