README.md

metadata

title: MuseControlLite
emoji: 👁
colorFrom: red
colorTo: blue
sdk: gradio
sdk_version: 6.1.0
app_file: app.py
pinned: false
license: mit
short_description: Inference for Stable-Audio-Open with more controls

MuseControlLite (Space)

Gradio UI for MuseControlLite adapters on top of stabilityai/stable-audio-open-1.0.

Requirements

• GPU Space is required for generation (fp16 by default).
• A Hugging Face token with access to stabilityai/stable-audio-open-1.0 (set as a Space secret, e.g., HF_TOKEN).

What happens on startup

1. Installs Python deps from requirements.txt (includes gradio, gdown, diffusers fork).
2. Downloads MuseControlLite checkpoints with
   gdown 1Q9B333jcq1czA11JKTbM-DHANJ8YqGbP --folder
   into checkpoints/ if they are missing.

Using the Space

1. Provide a text prompt.
2. Upload a 47.5s (or longer) audio file when using MuseControlLite conditions.
3. Select condition types (melody_stereo, melody_mono, dynamics, rhythm, audio) and adjust guidance/scales if needed.
4. Click Generate. Output is a single 47.5s WAV plus a short status summary.

Tips

• melody_stereo cannot be combined with dynamics, rhythm, or melody_mono.
• For audio in/out-painting, use the audio condition with the masking sliders.
• Default examples are preloaded in the UI for quick tests.

Local run (optional)

pip install -r requirements.txt
gdown 1Q9B333jcq1czA11JKTbM-DHANJ8YqGbP --folder
huggingface-cli login
python app.py

Acknowledgments

• Original repository: https://github.com/fundwotsai2001/MuseControlLite