EfficientSync achieves real-time lip sync by preserving reference textures

Researchers have developed EfficientSync, a novel real-time framework for audio-driven lip synchronization in videos. Unlike previous methods that reconstruct the lower face, EfficientSync focuses on faithfully transferring existing reference textures to maintain identity and authenticity. The system utilizes a Dynamic Texture Mixer for efficient multi-reference fusion, Spatio-Temporal Shifted Adaptive Masking to isolate lip generation conditions from the background, and STAR Sampling to select optimal reference frames. This approach achieves state-of-the-art visual quality and identity preservation at a high frame rate of 166 FPS on a single GPU. AI

IMPACT This research offers a more efficient and authentic approach to lip-syncing, potentially improving video conferencing and content creation tools.

RANK_REASON Academic paper detailing a new method and its experimental results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Dynamic Texture Mixer
• EfficientSync
• Hugging Face
• Spatio-Temporal Shifted Adaptive Masking
• STAR Sampling
• VFHQ

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →