Reproduce “Coding Speech through Vocal Tract Kinematics” 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
Cheol Jun Cho, Peter Wu, Tejas S. Prabhune, Dhruv Agarwal, Gopala K. Anumanchipalli
SPARC encodes ordinary speech as 14-channel articulatory and voice-source traces, then synthesizes speech with a separate speaker embedding. It estimates six articulators from audio and adds pitch and loudness. On LibriTTS-R test-clean, resynthesis has Whisper-large-v3 WER 5.43% and human MOS 3.82, versus 4.22% and 4.01, respectively, for original recordings. This supports near-preserved intelligibility and quality on this test, not universal performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.