# Audio Regenerate (UVM v2)

Research project page for **Audio Regenerate** — seamless, context-aware audio editing by generative latent inpainting. Powers word-level audio editing in [Descript](https://www.descript.com).

**Live site:** https://descriptinc.github.io/uvm-v2/

## Repo layout

| Path | Purpose |
|------|---------|
| `index.html` | Page structure |
| `styles.css`, `main.js` | Styling & interactions |
| `samples.js` | Generated sample data (do not hand-edit) |
| `build_samples.py` | Ingest impute output → `samples.js` + assets |
| `assets/samples/` | Audio + transcript files for the gallery |