Reproduce “Tap2Listen: Instant Target Switching for Streaming Audio-Visual Speaker Extraction” 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
Hansol Park, Minkyung Song, Zinu Kim, Yujun Lee, Hoseong Ahn, Hyunjin Choi, Kyuhong Shim
Tap2Listen extracts a speaker’s voice from streaming audio using the face. It keeps audio processing independent of face selection and recomputes decoder projections from cached audio after a tap. On 3,000 mixtures, it scores 9.68 dB SI-SNRi versus 8.85 dB for AV-SkiM. For switches to cached faces, it leads by 2.59 dB in the first 200 ms, not for unseen faces.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.