Vorch-Omni is presented, a unified multi-task framework for audio-visual synthesis based on an arbitrary-condition-to-arbitrary-output formulation that supports over 10 tasks, including text-to-video, text-to-audio-video, image- and reference-conditioned generation, temporal extension, audio-driven generation, video tr...
Vorch Team, Xiaoyu Chen, Yang Ding et al.· 0 citations
Recent advances in NeRF-based inpainting have enabled the completion of masked regions across multi-view images. However, two major challenges remain: generating accurate masks efficiently in the presence of complex multi-object interference and maintaining view consistency without floating artifacts in large-scale, un...