Video Diffusion Models
PulseAugur coverage of Video Diffusion Models — every cluster mentioning Video Diffusion Models across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
New frameworks leverage image editing to guide video diffusion models
Researchers have developed two new frameworks, EditDistill and VidEdit3D, to enhance video editing capabilities by leveraging image editing techniques and video diffusion models. EditDistill distills visual transformati…
-
WorldForge V2 enhances camera control in video diffusion models
Researchers have developed WorldForge V2, a novel framework designed to enhance camera trajectory control in video diffusion models. This method employs a training-free guidance approach that refines generations along a…
-
New method enables layout-guided video synthesis without additional training
Researchers have developed a novel training-free method for video synthesis that leverages existing video diffusion models to control object layout and motion. This approach operates in two stages: first, it uses prompt…
-
Video Diffusion Avatars create controllable 3D models from single images
Researchers have developed Video Diffusion Avatars (VDAs), a novel method for creating realistic, controllable 3D avatars from a single image. By fine-tuning video diffusion models, VDAs offer enhanced control over came…
-
GeoFix framework improves 4D video generation consistency
Researchers have developed GeoFix, a new framework designed to enhance the geometric consistency of 4D generated videos. This method utilizes a single geometric reward that focuses on cross-view depth consistency in sta…
-
AI Video Generation Better Aligns with Human Visual Cortex
Researchers have developed a new method to evaluate video generation models by comparing their internal representations to human visual cortex activity. They found that autoregressive (AR) video diffusion models, when g…
-
New research tackles video diffusion model efficiency and safety
Researchers are developing new methods to improve the efficiency and quality of video diffusion models. SplitMoE introduces a split-role architecture to better handle the semantic imbalance in video data, outperforming …
-
New research tackles reward model instability and efficiency in AI
Researchers are developing new methods to improve the training and robustness of reward models used in large language models and diffusion models. One approach, Density-Aware Reward Aggregation (DARA), addresses uneven …
-
DSAQuant framework improves video diffusion model quantization
Researchers have developed DSAQuant, a new framework for Quantization-Aware Training (QAT) specifically designed for Video Diffusion Models (VDMs). Existing QAT methods struggle with VDMs, often degrading visual details…
-
Video diffusion models scale effectively with training exposure, study finds
Researchers have conducted a comprehensive scaling-law analysis of video diffusion models specifically for autonomous driving applications. Their study, which involved models ranging from 1 million to 9 billion paramete…
-
AI advances tackle video generation consistency and speed
Recent research in AI-driven video generation focuses on improving consistency and efficiency. ContextAnyone addresses character consistency by treating reference images as explicit anchors within a shared diffusion tra…
-
DreamHand framework repurposes video diffusion models for 3D hand motion recovery
Researchers have developed DreamHand, a new framework that repurposes video diffusion models for improved 3D hand motion recovery from egocentric video. This method addresses challenges like object occlusion and hands l…
-
AvatarDynamizer transforms static 3D avatars into dynamic 4D versions
Researchers have developed AvatarDynamizer, a novel generative method designed to transform static 3D avatars into dynamic, realistic, and multi-view-consistent 4D avatars. This approach tackles the limitations of exist…
-
AnyTalk uses video diffusion models for 3D speech animation
Researchers have developed AnyTalk, a new method for creating 3D speech animations for arbitrary characters without needing existing animation data. This approach adapts pre-trained video diffusion models through a proc…
-
New video world model extrapolates physics laws with fewer parameters
Researchers have developed Latent Dynamics Reasoning (LDR), a novel approach to video world models that integrates kinematic dynamics into a structured latent space. This method allows models to extrapolate physical law…
-
New AI methods tackle video reflection removal and generation · 4 sources tracked
Researchers have developed novel methods for handling reflections in videos, addressing challenges in both removing unwanted reflections and generating realistic mirror reflections. One approach, S2R-Synthesis, uses phy…
-
New DAR method enables video models to render 4D scenes
Researchers have developed DAR, a novel approach that enables pretrained video diffusion models to function as 4D renderers. This method conditions these models using an animated mesh, camera trajectory, and a reference…
-
AgentHOI framework generates human-object interaction videos using multi-agent reasoning
Researchers have introduced AgentHOI, a novel framework for generating videos of human-object interactions. This system employs multi-agent reasoning to bridge the gap between textual descriptions and the physical execu…
-
MagicPrompt offers ultra-lightweight tuning for video generation
Researchers have introduced MagicPrompt, a novel framework designed to make video generation models more efficient. This method employs Attention-Embedded Prompt Tuning, which uses significantly fewer parameters than tr…
-
InverseCrafter framework enables efficient video re-capture without VDM fine-tuning
Researchers have introduced InverseCrafter, a novel framework for generating novel views of videos without requiring extensive fine-tuning of pre-trained Video Diffusion Models (VDMs). This approach treats video re-capt…