Tag
UniSwap is a new framework for joint audio-visual identity swapping in talking videos, using a unified streaming audio-visual diffusion transformer to replace appearance and vocal timbre while preserving source content and dynamics.