ACE-Step
Open-source music generation foundation model delivering SOTA performance
AI / MLProduct
About ACE-Step
Open-source music generation, state-of-the-art results
ACE-Step is an open-source foundation model for music generation. A single text prompt produces complete songs with vocals and instrumentals, with generation speed far beyond diffusion-based baselines and quality that holds up against closed models.
For developers and creators
Self-host the model, fine-tune it on your own style and data, or try the hosted demo at ace-step.club. Weights are published on Hugging Face, the source code is on GitHub, and an active Discord community shares checkpoints and workflows.
Key features
- Text-to-song generation with vocals and instrumentals
- Open-source weights, self-hosting and fine-tuning
- Hosted demo at ace-step.club
