Reproduce “Vocal Intensity Conversion with Continuous SPL Control via Adversarial Training” 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
Quentin Le Tellier, Albert Rilliard, Olivier Perrotin, Marc Evrard
The authors introduce a speech conversion system that conditions a frozen analysis-and-resynthesis pipeline on a continuous sound-pressure-level target in decibels. A trainable feature converter uses no matched target recording. On WavLM features, predicted intensity displacement has a regression slope of 0.83, while listening tests show perceived level rises with target, but its absolute level remains source-dependent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.