Reproduce “ARTI-6: Towards Six-dimensional Articulatory Speech Encoding” 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
Jihwan Lee, Sean Foley, Thanathai Lertpetchpun, Kevin Huang, Yoonjeong Lee, Tiantian Feng, Louis Goldstein, Dani Byrd and 1 more
ARTI-6 turns six regions from real-time MRI into interpretable signals, predicts them from speech, and uses them to synthesize speech. On a single-speaker test set, WavLM-Large predicts the features with a Pearson correlation of 0.872. On LibriTTS-R, synthesis has 12.5% WER, versus 4.8% for a 14-dimensional EMA, pitch, and loudness baseline, showing a measured intelligibility gap.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.