[ECCV 2024] Enhancing Perceptual Quality in Video Super-Resolution through Temporally-Consistent Detail Synthesis using Diffusion Models
-
Updated
Feb 7, 2025 - Python
[ECCV 2024] Enhancing Perceptual Quality in Video Super-Resolution through Temporally-Consistent Detail Synthesis using Diffusion Models
[NeurIPS`25] TC-Light: Temporally Coherent Generative Rendering for Realistic World Transfer
DeepDream for video with temporal consistency. Features RAFT optical flow estimation and occlusion masking to prevent ghosting. A PyTorch implementation.
Add Temporal Coherence to AI Upscales in VapourSynth
This repo contains the official implementation of ICCV 2025 paper "MoSiC: Optimal-Transport Motion Trajectory for Dense Self-Supervised Learning
[IEEE Signal Processing Letters, 2025] Adaptive Video Demoiréing Network with Subtraction-Guided Alignment
Implementation of the paper DynaMesh: Dynamic 3D Texture Generation.
Few-step diffusion for audio-driven talking head generation making diffusion models speak faster without losing their composure.
Progressive-resolution sampling for ComfyUI: run the early denoising steps at low resolution, lift the result to full resolution, and finish there — faster generation with no training.
Optical flow for Houdini 22 Copernicus. Where each pixel moved, and whether to trust it.
Domain-specific restoration pipeline for broadcast anime video on the Re-Anime600 benchmark (AINIME Challenge @ WACV 2027 Workshop).
MiniMax H3 video continuation for Wan2GP. Watch the native reframe disappear, then see a 40-second, 3-window result. Overlap replay, bounded video state, real before/after demos, and the full method.
Official Python implementation for AIRC-Fusion: Proving that real-time IR-VIS video fusion is possible on standard CPUs (tested on Raspberry Pi 5)
Production-scale video style transfer (AdaIN + RAFT Optical Flow) achieving 6.45 FPS and trained via DDP on 118K images.
Code release for EVEWorld: physical evolution supervision for target-instance and cross-frame consistency in embodied world models.
Repository for implementing Temporal Consistency in Semantic Segmentation model
ComfyUI nodes for Blender EXR depth/normal passes → ControlNet temporal conditioning
LTX and Hunyuan video post-training with sealed-holdout evaluation
Real-time generative rendering engine: grey-box + reference + first-frame + prompt to photoreal/stylized frames at 24 FPS (zero-dependency reference of VAST EdenLive).
[IEEE SPL 2024] A RNN for temporal consistency in videos enhanced by single-frame methods
To associate your repository with the temporal-consistency topic, visit your repo's landing page and select "manage topics."