Newest papers

From Short Clips to Real-Time Streams: Identity-Aware Supervision and Context-Efficient Generation for Video Face Swapping

Zekai Luo, Yuwei Liang, Ruiqi Ma, Haoyu Huang, Xuehao Wang, Huanyi Zheng, Mingyu Liu, Linhao Zhong, Chenchen Jing, Yuling Xi, Hao Chen · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26681 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Pertukaran wajah video tunggal mentransfer identitas referensi ke video target sambil mempertahankan atribut dan gerakan lainnya. Terlepas dari kemajuan baru-baru ini dalam metode berbasis arus referensi, bersama-sama mencapai transfer identitas yang setia, loyalitas tinggi, dan generasi real-time di atas aliran video panjang tetap menantang. Untuk mengatasi tantangan ini, kami mempertimbangkan kembali struktur spesifik identitas eksplisit dan rekonstruksi latent sesuai dengan tingkat kebisingan. Pertama, video target bertukar hanya identitas sementara mempertahankan sebagian besar konten visual, namun pengawasan flow-matching sangat sensitif terhadap perubahan ini. Oleh karena itu kami mengembangkan Joint Rec-ID Loss, yang menyeimbangkan pengawasan identitas eksplisit dan rekonstruksi latent sesuai dengan tingkat kebisingan. Kedua, video target menyediakan atribut yang sudah tidak ditampilkan dan dinamika

Machine translation (Meta M2M100); the English original follows.

Abstract

Video face swapping transfers the reference identity to a target video while preserving its other attributes and motion. Despite recent progress in reference-guided flow-based methods, jointly achieving faithful identity transfer, high fidelity, and real-time generation over long video streams remains challenging. To address this challenge, we rethink the task-specific structure of reference-guided video face swapping. First, face swapping changes only identity while preserving most visual content, yet flow-matching supervision is weakly sensitive to these changes. We therefore develop Joint Rec-ID Loss, which balances explicit identity supervision and latent reconstruction according to the noise level. Second, the target video already provides non-edited attributes and temporal dynamics, enabling generation with fewer spatial tokens and shorter temporal contexts through Face-RoI Generation and Short-Context Rollout Training. By integrating these designs, we propose StreamingSwap, a reference-guided flow-matching framework for real-time streaming video face swapping. We also introduce HumanFaceVid, comprising 1,000 hours of cinematic training data and evaluation sets of 200 short and 200 long videos. Experiments demonstrate that our method achieves competitive identity transfer and visual quality at 20 FPS on a single NVIDIA RTX 4090 GPU, with stable generation over videos up to 10 minutes long. Project webpage: https://aim-uofa.github.io/StreamingSwap/

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Identity (music), Face (sociological concept), Artificial intelligence, Computer vision, Quality (philosophy)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange