Research project page for Audio Regenerate — seamless, context-aware audio editing by generative latent inpainting. Powers word-level audio editing in Descript.
Live site: https://descriptinc.github.io/uvm-v2/
| 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 |