Tagged articles

temporal consistency

4 articles · Page 1 of 1

Sep 5, 2026 · Artificial Intelligence

Designing an AI Face-Swap System for Precise Expression Transfer: 3DMM + Diffusion Pipeline

The article details an end-to-end AI face-swap system using YOLOv11-Face detection, FLAME 2.0 with neural refinement for 3D expression disentanglement, Stable Diffusion XL with ControlNet and IP-Adapter for generation, and post-processing for blending, lighting, and temporal consistency, achieving near 30fps on RTX 4090 with compliance measures.

3DMMAI face swapControlNet

0 likes · 14 min read

Designing an AI Face-Swap System for Precise Expression Transfer: 3DMM + Diffusion Pipeline

Kuaishou Tech

Jun 8, 2026 · Artificial Intelligence

CVPR NTIRE 2026 UGC Short Video Restoration Challenge: Winning Solutions Revealed

The CVPR NTIRE 2026 challenge introduced the KwaiVIR benchmark for real‑world UGC short‑video degradation and hosted two tracks—subjective and objective—where 12 teams submitted results, with RedMediaTech achieving top scores across fidelity, perceptual quality, and temporal consistency, followed by detailed analyses of the leading methods.

KwaiVIR datasetNTIRE 2026UGC video restoration

0 likes · 10 min read

CVPR NTIRE 2026 UGC Short Video Restoration Challenge: Winning Solutions Revealed

DataFunTalk

Jan 25, 2024 · Artificial Intelligence

Hierarchical Masked 3D Diffusion Model for Video Outpainting

This paper introduces a hierarchical masked 3D diffusion model (M3DDM) that leverages mask modeling and global-frame cross‑attention to achieve temporally consistent video outpainting, proposes a hybrid coarse‑to‑fine inference pipeline to mitigate error accumulation in long videos, and demonstrates state‑of‑the‑art results on benchmark datasets.

3D U-NetAItemporal consistency

0 likes · 13 min read

Hierarchical Masked 3D Diffusion Model for Video Outpainting

Alimama Tech

Jan 24, 2024 · Artificial Intelligence

Hierarchical Masked 3D Diffusion Model for Video Outpainting

The Hierarchical Masked 3D Diffusion Model (M3DDM) introduces a masking‑based training strategy and cross‑attention with global video clips to achieve temporally consistent video outpainting, while a hybrid coarse‑to‑fine inference pipeline mitigates error accumulation, delivering state‑of‑the‑art results and deployment in Alibaba’s creative center.

3D diffusionACM MM2023AI video processing

0 likes · 12 min read