Colab Wrappers for HarmonAI’s Dance Diffusion

What it does This repository hosts Jupyter notebooks for running Dance Diffusion, HarmonAI’s generative audio model. It provides Colab-ready environments for inference, fine-tuning, and sample interpolation on whatever audio clips you supply. The whole affair is essentially minimalist scaffolding around the underlying diffusion model.

The interesting bit The README treats exposition as someone else’s problem: no architecture diagrams, no audio examples, no tutorials—just notebook badges and a four-item todo list. For better or worse, the project trusts that you already speak diffusion.

Key highlights

• Ready-to-run Colab notebooks for inference and fine-tuning
• Supports interpolations between generated samples
• Accepts arbitrary audio samples for custom training
• Public todo list shows guidance features are still pending

Caveats

• README is extremely sparse; newcomers will need external context to get oriented
• A typo in the todo list (“nobebook”) suggests maintenance is casual
• The guidance notebook promised in the todo remains unchecked

Verdict Useful if you are already experimenting with HarmonAI’s diffusion models and want ready-made training environments. Everyone else should read up on Dance Diffusion first, because this repo will not explain it to you.