Reproduce “FoleyForge: Free-form, Controllable, Rectified-flow Editing of Sound Effects” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Giovana Morais, Hugo Flores García, Prem Seetharaman, Johannes Imort, Justin Salamon, Magdalena Fuentes, Oriol Nieto
FoleyForge edits existing sound clips from free-form instructions while conditioning on source audio, using clean source latent tokens prepended to noisy output tokens and adjustable initialization noise to trade source faithfulness for adherence. On a 90-input Creative Set, its Qwen instruction-adherence rating was 3.44/5, versus 2.69/5 for SAO-Instruct; this judge-rated result is not a universal win or human rating.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.