audio-studio

Local music generation studio wrapping ACE-Step 1.5.

Text-to-music, reference-audio melody borrowing, and reference-audio cover generation, with configurable artist/album/track/genre texture hints.

Features

  • Text-to-music: Describe what you want, get a WAV file
  • Melody borrowing: Provide a reference audio to borrow its BPM, key, and duration
  • Cover mode: Full sound/texture conditioning from a reference (requires ace-step backend)
  • Texture hints: Artist, album, track, and genre hints appended to prompts
  • Two backends: mlx-serve (default, 2.9GB) or ace-step (17.8GB, for cover mode)

Quick Start

# Clone and setup
git clone https://github.com/sm3222/audio-studio.git
cd audio-studio
uv sync

# Start the backend ( mlx-serve by default)
mlx-serve serve --port 8082 --model-dir ~/.mlx-serve/models

# Generate music
uv run audio-studio "a calm piano piece" --artist "Bob Marley" --output out.wav

Links

Backend
Example Prompts