---
格式版本: 2
标题: "EfficientSync: Real-Time Lip Synchronization via Deformation-Based Reference Texture Mixing"
原文链接: "https://arxiv.org/abs/2608.18832"
发布日期: "2026-08-19"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Wed, 19 Aug 2026 11:49:06 UTC (3,788 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-20T15:35:13+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-20T15:33:55+08:00"
入库时间: "2026-08-20T07:35:13.214Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=latency&searchtype=all"
匹配关键词:
  - "latency"
  - "GPU"
相关厂家:
  []
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 8
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文内容为唇形同步技术，与超节点/AI Rack等主题完全无关，仅命中检索词latency，无任何相关技术或商业信息。"
AI质检模型: "tx-deepseek-v4-flash"
AI质检时间: "2026-08-20T15:38:13+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 3
AI摘要: "EfficientSync提出一种基于变形参考纹理混合的实时唇形同步框架，通过动态纹理混合器以通道加权方式保留参考纹理，并用自适应掩码分离唇部与背景，避免重生成导致的口腔细节失真。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:19:20.438Z"
采集批次: "2026年8月20日14点19分32秒"
采集批次ID: "20260820-141932-079"
去重键: "https://arxiv.org/abs/2608.18832"
---

## Computer Science > Computer Vision and Pattern Recognition

## Title:EfficientSync: Real-Time Lip Synchronization via Deformation-Based Reference Texture Mixing

Authors:[Fa-Ting Hong](https://arxiv.org/search/cs?searchtype=author&query=Hong,+F), [Runzhen Liu](https://arxiv.org/search/cs?searchtype=author&query=Liu,+R), [Luchuan Song](https://arxiv.org/search/cs?searchtype=author&query=Song,+L), [Hongmin Cai](https://arxiv.org/search/cs?searchtype=author&query=Cai,+H), [Chuhua Xian](https://arxiv.org/search/cs?searchtype=author&query=Xian,+C)

[View PDF](https://arxiv.org/pdf/2608.18832) [HTML (experimental)](https://arxiv.org/html/2608.18832v1)

> Abstract:Audio-driven lip synchronization manipulates the mouth region of a talking-face video to match the driving audio while preserving head pose, identity, and background. Although the task is inherently local editing, prevailing approaches reconstruct the entire lower face with heavy GAN- or diffusion-based decoders, incurring substantial latency and, more critically, hallucinating intra-oral details such as teeth and lip wrinkles instead of preserving authentic textures. We contend that the bottleneck in identity preservation is not the scarcity of reference frames, but the lack of a mechanism that faithfully transfers the genuine textures they already contain. We therefore present EfficientSync, a real-time deformation-based framework that retains reference textures rather than resynthesizing them. First, the Dynamic Texture Mixer reformulates multi-reference fusion as channel-wise selection, evaluating each spatially aligned reference in a global context and aggregating them by channel-wise weighted summation, preserving textural integrity at low cost. Second, Spatio-Temporal Shifted Adaptive Masking decomposes the source frame into lip-generation conditions and an independent background prior, suppressing lower-face leakage while blending the synthesized mouth seamlessly into the background. Third, STAR Sampling, a zero-overhead pre-processing step, retrieves the sharpest and most topologically diverse reference frames. Experiments on HDTF and VFHQ show state-of-the-art visual quality and identity preservation at 166 FPS on a single GPU. Video demos: [this https URL](https://alunaticat.github.io/EfficientSync/index.html).

| Comments: |  |
| --- | --- |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | [arXiv:2608.18832](https://arxiv.org/abs/2608.18832) \[cs.CV\] |
|  | (or [arXiv:2608.18832v1](https://arxiv.org/abs/2608.18832v1) \[cs.CV\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.18832](https://doi.org/10.48550/arXiv.2608.18832) |

## Submission history

From: Fa-Ting Hong \[[view email](https://arxiv.org/show-email/f82e8a1e/2608.18832)\]  
**\[v1\]** Wed, 19 Aug 2026 11:49:06 UTC (3,788 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.18832) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))