#VideoDiffusion
LightNews LightNews
aidailypost.com
🚀 Nunchux AI just dropped VC‑Attention, a training‑free low‑bit kernel that speeds up video Diffusion Transformers. Faster video generation, no extra training, and a softmax boost. Curious? Dive in! #VCAttention #NunchuxAI #VideoDiffusion

🔗 aidailypost.com/news/nunchux...
September 17, 2026 at 1:04 AM
feiticeir0s.bsky.social
How I Created Stunning AI Videos Locally, Free, Offline, and Uncensored.
#medium #FramePackt #ai #aivideo #aivideodiffusion #videodiffusion

github.com/lllyasviel/F...

If this works well, this could be something very interesting.
How I Created Stunning AI Videos Locally, Free, Offline, and Uncensored
From static photos to fully animated scenes, here’s how I turned my laptop into a cinematic AI video studio, and how you can too.
ai.plainenglish.io
May 15, 2025 at 8:08 AM
pneumetron.bsky.social
🤖 4DAnyone Solves the Scaling Bottleneck in 4D Human Reconstruction

https://pneumetron.com/news/ai_research/4danyone-solves-scaling-bottleneck-4d-human-reconstruction-21fe2f

#computervision #4dgaussiansplatting #videodiffusion #generativeai
August 26, 2026 at 6:37 AM
getnews-me.bsky.social
ReLumix’s two‑stage pipeline relights a reference frame via any image‑relighting tool, then a stable video diffusion model spreads the lighting with temporal coherence. Read more: https://getnews.me/relumix-video-relighting-via-image-techniques-and-diffusion-models/ #relighting #videodiffusion
ReLumix: Video Relighting via Image Techniques and Diffusion Models
September 30, 2025 at 2:02 PM
getnews-me.bsky.social
QuantSparse reduces storage 3.68× and inference time 1.88×, achieving 20.88 dB PSNR on HunyuanVideo‑13B, ahead of Q‑VDiT's 16.85 dB. https://getnews.me/quantsparse-faster-video-diffusion-transformers-with-sparse-quantization/ #quantsparse #videodiffusion
QuantSparse: Faster Video Diffusion Transformers with Sparse Quantization
September 30, 2025 at 11:51 AM
getnews-me.bsky.social
Astraea speeds up video diffusion transformers, delivering up to 2.4× faster inference on a single GPU with under 0.5% loss in VBench quality metrics. Read more: https://getnews.me/astraea-framework-accelerates-video-diffusion-transformers/ #videodiffusion #ai
Astraea Framework Accelerates Video Diffusion Transformers
September 29, 2025 at 10:31 PM
getnews-me.bsky.social
A new spectral loss for video diffusion models improves action‑recognition by ~11% and cuts warping error up to 37%, using only 2.7% of frequency coefficients. June 2025. https://getnews.me/physics-guided-motion-loss-boosts-realism-in-video-generation-models/ #videodiffusion #physics
Physics-Guided Motion Loss Boosts Realism in Video Generation Models
September 29, 2025 at 10:24 PM
getnews-me.bsky.social
PhyMAGIC animates a single image untrained, using diffusion and physics simulation, and outperforms video diffusion models on physical consistency. Read more: https://getnews.me/phymagic-training-free-physical-motion-generation-with-llm-guidance/ #phymagic #videodiffusion
PhyMAGIC: Training-Free Physical Motion Generation with LLM Guidance
September 29, 2025 at 9:36 PM
getnews-me.bsky.social
DiTraj is a training‑free method that uses an LLM to split prompts into foreground and background and applies STD‑RoPE to adjust position embeddings for smoother motion. Read more: https://getnews.me/ditraj-training-free-trajectory-control-for-video-diffusion-transformers/ #ditraj #videodiffusion
DiTraj: Training‑Free Trajectory Control for Video Diffusion Transformers
September 29, 2025 at 9:17 AM
getnews-me.bsky.social
A new autoregressive video diffusion model streams video at 9.4 fps on a single GPU, achieving an 84.27 VBench‑Long score and cutting inference from 50 to 4 steps. https://getnews.me/fast-autoregressive-video-diffusion-models-boost-real-time-generation/ #videodiffusion #realtime
Fast Autoregressive Video Diffusion Models Boost Real-Time Generation
September 27, 2025 at 12:32 AM
getnews-me.bsky.social
BWCache, a training‑free technique, cuts video diffusion transformer inference time up to 2.24× by caching block features and reusing them when changes are minimal. Read more: https://getnews.me/bwcache-accelerates-video-diffusion-transformers-via-block-wise-caching/ #videodiffusion #bwcache
BWCache Accelerates Video Diffusion Transformers via Block‑Wise Caching
September 20, 2025 at 2:56 PM
getnews-me.bsky.social
WorldForge unveiled a training‑free plug‑in that adds precise motion control to video diffusion models, enabling realistic 3D/4D content without retraining. Submitted September 18, 2025. https://getnews.me/worldforge-launches-training-free-3d-4d-video-generation/ #worldforge #videodiffusion
WorldForge Launches Training-Free 3D/4D Video Generation
September 20, 2025 at 3:11 AM
kinews24.bsky.social
Stability AI präsentiert Stable Video 3D: Qualitativ hochwertige Synthese neuer Ansichten und 3D-Generierung aus Einzelbildern

#KuenstlicheIntelligenz #artificialintelligence #KI #AI #3DGenerierung #Videodiffusion #StableVideo3D #ComputerVision

kinews24.de/stability-ai...
March 19, 2024 at 5:44 PM
arxivlens.bsky.social
Diffusion Knows Transparency: Repurposing Video Diffusion for Transparent Object Depth and Normal Estimation
Bin Ma, Bohan Li et al.
Paper
Details
#VideoDiffusion #DepthEstimation #ComputerVision
December 30, 2025 at 9:02 AM
arxivlens.bsky.social
AnchorDream: Repurposing Video Diffusion for Embodiment-Aware Robot Data Synthesis
Basile Van Hoorick, Junjie Ye et al.
Paper
Details
#EmbodimentAware #RobotDataSynthesis #VideoDiffusion
December 15, 2025 at 9:03 AM
arxivlens.bsky.social
Video2Act: A Dual-System Video Diffusion Policy with Robotic Spatio-Motional Modeling
Boxin Shi, Jiaming Liu et al.
Paper
Details
#VideoDiffusion #Robotics #SpatioMotionalModeling
December 3, 2025 at 11:46 AM
arxivlens.bsky.social
RAP: Real-time Audio-driven Portrait Animation with Video Diffusion
Transformer
Dingcheng Zhen, Fangyu Du et al.
Paper
Details
#RAPAnimation #VideoDiffusion #TransformerTech
August 10, 2025 at 4:02 PM