stable-audio-tools

Stability AI’s training and fine-tuning toolkit for their audio generation models, with a Gradio interface.

Strengths

• The training path for Stable Audio Open — fine-tune on your own sample library
• MIT-licensed code, actively maintained
• Gradio UI for auditioning models without writing code
• Model unwrapping turns a training checkpoint into a usable model
• Supports Stable Audio Open Small for on-device work

Limitations

• The weights you train from carry Stability’s Community License, not MIT
• Research-oriented: the API moves faster than diffusers
• Requires PyTorch 2.5+ for Flash Attention — a very different pin from AudioCraft, so use separate environments
• For plain inference, diffusers is less work

Best for

Fine-tuning Stable Audio Open on your own audio, or running the Gradio UI against the released weights.

Minimal example

pip install stable-audio-tools

# Gradio UI against the released weights
python3 ./run_gradio.py --pretrained-name stabilityai/stable-audio-open-1.0

# Fine-tune on your own dataset
python3 ./train.py --dataset-config /path/to/dataset.json \
                   --model-config /path/to/model.json \
                   --name my_finetune

Related tools

See all stable-audio-tools alternatives & how they compare →

See the whole ML & generative category →

← Back to all tools