Skip to main content
GameDev.net gamedev.net

Video Generation News

The latest Video Generation coverage curated for game developers.

PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control

arXiv cs.GR details PhysStream, a video-generation model that lets creators steer physics-driven motion mid-generation instead of locking in a full control schedule up front. For game teams, the interesting bit …

Research arXiv cs.GR · 1 week, 1 day ago
EditStream: A Unified Autoregressive Framework for Interactive Video Generation and Editing

EditStream folds text-to-video, image-to-video, video-to-video, editing propagation, reference-guided edits, and camera pose changes into one DiT-based system. The big shift for developers is its few-step autoregressive streaming setup, aimed at …

Research arXiv cs.GR · 4 weeks, 2 days ago
WorldRover: A Scalable Synthetic Video Data Engine for World Exploration with Rich Annotations

WorldRover turns long-horizon world exploration into a synthetic data problem, using Unreal Engine to render minute-scale traversals with full trajectories and scene geometry. The resulting 10M-sequence dataset pairs RGB with …

Research arXiv cs.GR · 1 month ago
KeyID: Decoupled Drafting and Keyframe Editing for Identity-Preserving Video Generation

KeyID tackles identity-preserving video generation by splitting motion synthesis from identity injection, cutting the usual tug-of-war between prompt fidelity and subject consistency. The training-free pipeline drafts an identity-agnostic sequence first, …

Research arXiv cs.GR · 1 month ago
Wonder: Video World Model Done Better

Wonder is a new video world model that turns a single image or conditioned clip into a playable scene you can explore by moving the camera in real time. The …

Research arXiv cs.GR · 1 month, 3 weeks ago
Echo-Infinity: Learning Evolving Memory for Real-Time Infinite Video Generation

A new AR video-generation paper claims real-time rollouts for more than 24 hours, crossing 1.3 million frames without the usual cache-window collapse. The interesting bit for game teams is the …

Research arXiv cs.GR · 3 months, 2 weeks ago
3DPhysVideo: Consistency-Guided Flow SDE for Video Generation via 3D Scene Reconstruction and Physical Simulation

A new training-free pipeline claims to turn a single image into physically plausible video by first reconstructing 360° scene geometry, then running physics on that geometry before reusing the same …

Research arXiv cs.GR · 4 months ago
Delta Forcing: Trust Region Steering for Interactive Autoregressive Video Generation

A new video-generation method called Delta Forcing aims to fix a familiar problem for interactive AI systems: staying responsive without drifting off-model over time. It uses a trust-region style constraint …

Research arXiv cs.GR · 4 months, 1 week ago
Alice v1: Distillation-Enhanced Video Generation Surpassing Closed-Source Models

Alice v1 introduces a groundbreaking 14-billion parameter open-source video generation model that outperforms closed-source counterparts. This model is particularly relevant for graphics programmers and artists, as it achieves state-of-the-art video …

Research arXiv cs.GR · 4 months, 1 week ago
SURF: Signature-Retained Fast Video Generation

SURF tackles a practical bottleneck for teams experimenting with video generation: 720p inference can take 50+ minutes on Wan2.1. The paper’s two-stage approach keeps more of the base model’s layout, …

Research arXiv cs.GR · 6 months ago
Physical Simulator In-the-Loop Video Generation

The introduction of Physical Simulator In-the-Loop Video Generation (PSIVG) marks a significant leap in video generation technology for game developers. By integrating a physical simulator into the video diffusion process, …

Research arXiv cs.GR · 6 months, 2 weeks ago
RealWonder: Real-Time Physical Action-Conditioned Video Generation

RealWonder introduces a groundbreaking approach to video generation by integrating real-time physics simulation with action conditioning. This innovation allows developers to create interactive experiences that simulate physical consequences in 3D …

Research arXiv cs.GR · 6 months, 2 weeks ago
Adaptive Hybrid Caching for Efficient Text-to-Video Diffusion Model Acceleration

MixCache introduces a novel caching framework that significantly boosts video generation speed, achieving up to 1.97x acceleration. This is particularly beneficial for graphics programmers and artists focused on multimedia content …

Research arXiv cs.GR · 6 months, 4 weeks ago
FlexAM: Flexible Appearance-Motion Decomposition for Versatile Video Generation Control

A new video-generation control framework separates appearance from motion using a 3D point-cloud signal, aiming to make editing more reliable across tasks. FlexAM targets image-to-video, video-to-video, camera control, and spatial …

Research arXiv cs.GR · 7 months, 1 week ago
FreeOrbit4D: Training-Free Arbitrary Camera Redirection for Monocular Videos via Foreground-Complete 4D Reconstruction

A training-free video redirection pipeline is trying to solve a nasty problem for camera-driven content: how to move far from the original monocular view without the scene falling apart. FreeOrbit4D …

Research arXiv cs.GR · 7 months, 3 weeks ago
MV-S2V: Multi-View Subject-Consistent Video Generation

The MV-S2V framework introduces a significant leap in video generation by synthesizing content from multiple views, enhancing 3D subject consistency. This advancement is particularly relevant for graphics programmers and artists, …

Research arXiv cs.GR · 7 months, 4 weeks ago
Lightning Fast Caching-based Parallel Denoising Prediction for Accelerating Talking Head Generation

The introduction of LightningCP significantly enhances the efficiency of talking head generation, making it a game changer for graphics programmers. By caching static features, it reduces inference time, allowing for …

Research arXiv cs.GR · 8 months ago
Controllable Video Generation: A Survey

Recent advancements in controllable video generation are reshaping how developers can create content that aligns closely with user intent. By integrating non-textual conditions like camera motion and depth maps, these …

Research arXiv cs.GR · 8 months ago
Beyond Inpainting: Unleash 3D Understanding for Precise Camera-Controlled Video Generation

DepthDirector introduces a new framework for video generation that enhances camera control while preserving video content fidelity. This advancement is particularly relevant for graphics programmers and artists, as it addresses …

Research arXiv cs.GR · 8 months, 1 week ago
ByteLoom: Weaving Geometry-Consistent Human-Object Interactions through Progressive Curriculum Learning

ByteLoom introduces a novel approach to human-object interaction in video generation, leveraging a Diffusion Transformer framework. This innovation addresses critical challenges in cross-view consistency and reduces reliance on detailed hand …

Research arXiv cs.GR · 8 months, 3 weeks ago
Page 1 of 2 Next